|
|
Post by ansat on Jan 20, 2015 16:59:40 GMT -5
Those who did a rew comparison of the emotiva microphone vs another calibrated mic, please PM me. I have a new set of calibration files that I need tested.
|
|
|
|
Post by barrak on Jan 20, 2015 19:41:21 GMT -5
This might be of interest regarding the mic sensitivity at the upper end. The following are measurements of my center speaker, which has a concentric tweeter/midrange arrangement. The first plot is with the speaker laying flat with the tweeter at 14" from the floor. The second, with the speaker tilted to face the intitial mic location (9' away). One would think that the upper end of the measurement would rise a bit. However, both plots are nearly identical up there.   |
|
tubby
Emo VIPs 
Route 2 in Weekapaug!!!
Posts: 408
|
Post by tubby on Jan 20, 2015 20:07:43 GMT -5
This might be of interest regarding the mic sensitivity at the upper end. The following are measurements of my center speaker, which has a concentric tweeter/midrange arrangement. The first plot is with the speaker laying flat with the tweeter at 14" from the floor. The second, with the speaker tilted to face the intitial mic location (9' away). One would think that the upper end of the measurement would rise a bit. However, both plots are nearly identical up there. View AttachmentView AttachmentSingle point or 9 point measurement? |
|
|
|
Post by barrak on Jan 20, 2015 20:31:43 GMT -5
This might be of interest regarding the mic sensitivity at the upper end. The following are measurements of my center speaker, which has a concentric tweeter/midrange arrangement. The first plot is with the speaker laying flat with the tweeter at 14" from the floor. The second, with the speaker tilted to face the intitial mic location (9' away). One would think that the upper end of the measurement would rise a bit. However, both plots are nearly identical up there. Single point or 9 point measurement? Full measurement suite for both... 9 measurements. |
|
|
|
Post by socketman on Jan 20, 2015 20:37:20 GMT -5
Those who did a rew comparison of the emotiva microphone vs another calibrated mic, please PM me. I have a new set of calibration files that I need tested. I PM'd you 5 or 6 times  |
|
tubby
Emo VIPs
Route 2 in Weekapaug!!!
Posts: 408
|
Post by tubby on Jan 20, 2015 20:46:13 GMT -5
Single point or 9 point measurement? Full measurement suite for both... 9 measurements. The averaging may account for the minimal change. |
|
|
|
Post by barrak on Jan 20, 2015 21:02:26 GMT -5
Full measurement suite for both... 9 measurements. The averaging may account for the minimal change. Possibly, though there was relative smoothing between 300 Hz and 3 kHz after tilting the speaker. The midrange response was affected, but not the tweeter's! I guess it could be that this particular 8" midrange cone has a narrow beam at its upper range, compared to the tweeter. |
|
|
|
Post by socketman on Jan 20, 2015 21:59:25 GMT -5
|
|
|
|
Post by ansat on Jan 20, 2015 22:00:36 GMT -5
Ok - So I have been seeing 3 different trends in the files posted here or the files sent to me in private. There is the everything looks like a big hump graphs. The graphs with the issues starting at 2khz - 3khz and having a slight drop-off (6db) and the 2khz - 3khz having a major drop-off around 10k (up to 12db). The big hump -- Its either the speaker or the microphone. Look up some graphs of what you speaker should look like or talk to the manufacturer of your speaker and explain what you are seeing. If you suspect that the speaker is right -- call emotiva, you likely have a bad microphone. For the other two. I have looked at a few microphone comparison files and I have come up with 2 different calibrations. One is just adding the differences between a 90* and 0* to Emotiva's existing calibration (5db correction) as well as fixing the lower end that is the same on all but one file that I reviewed. (6db starting at 100hz) My particular microphone also had an issue between 300 and 800 hz (3db) which I corrected in addition to the above corrections and microphone did not end at 5db off at 20khz  I would recommend that the math cal be used as it should fit the most microphones out there. However, I have also included the calibration that is matched to my microphone. My cal filesto use the calibration - you simply need to close Dirac. replace the calibration file found at C:\Users\YOURUSERNAME\AppData\Roaming\Dirac\OEM\Emotiva\MicCalFiles (windows 8) with one of mine. As soon as these files are no longer needed (Emotiva/Dirac makes the corrections)(Or give me something to show that I am just imagining things) I will remove the files from my server and remove this post(maybe the entire thread). Tony |
|
|
|
Post by solidstate on Jan 20, 2015 23:30:35 GMT -5
SS, I was wondering when you were going to join the party. I am on my mobile right now, I will send you my file in the morning. Thanks bro! |
|
|
|
Post by socketman on Jan 20, 2015 23:30:53 GMT -5
What I am not sure of is if this is a Dirac software issue or a Mic issue separate from the cal file. I believe it would be prudent for Dirac to allow for the use of other microphones.
|
|
|
|
Post by markc on Jan 21, 2015 1:54:27 GMT -5
This might be of interest regarding the mic sensitivity at the upper end. The following are measurements of my center speaker, which has a concentric tweeter/midrange arrangement. The first plot is with the speaker laying flat with the tweeter at 14" from the floor. The second, with the speaker tilted to face the intitial mic location (9' away). One would think that the upper end of the measurement would rise a bit. However, both plots are nearly identical up there. Single point or 9 point measurement? I think the issue is that none of us should believe that a centre speaker (limited volume cabinet, possibly smaller drivers, internal crossovers optimised for dialogue clarity etc) could possibly have such AMAZING low end output (from 20Hz up) and yet a massive roll off starting at 5kHz.
Speakers are just not made this way and that is why we use subs and why a sub woofer is usually run with a crossover between 50-120Hz to compensate for the inability of standard speakers to reproduce these frequencies and why we only need one tweeter irrespective of the number of mid/low range drivers in the cabinet - they can just produce the correct high frequencies easily.
We are not measuring what we think we are measuring and it is a microphone problem or a calibration file problem or possibly something going on in Windows / Dirac. Worse, we are believing what we are seeing on screen when it cannot possibly be correct and debating plausible workarounds to modify it.
Unless the calibration file for the Emo mic has some incorrect and major positive corrections at the low end and similarly aggressive negative corrections for the high frequencies, then the mic itself is not measuring correctly.
I have only ever used an analogue mic into soundcard for REW rather than the USB interface. Could this be somewhere that a problem lies rather than the EMO mic?
Is it the test tone? Is it incorrect or not what Dirac is expecting?
Here is a picture from the UMIK page as their example of how a speaker should measure in Dirac, which shows what we would all expect a full range speaker to look like, not like ANY of the screenshots of XMC users. The trace climbs up to 50/60Hx and is pretty steady until a roll off starting 17kHz.
www.minidsp.com/images/appnotes/umik-1-dirac/umik-1-dirac-measured.png 
~Here is a picture from the Dirac website. It also show us what we know a speaker response should look like. Acceptable performance from 50-20kHz with only the expected and usual variations across the trace due to nulls/voids/reflections etc 
Here is someone's REW measurement using a UMIK mic from the REW forums. This is the same pattern as how my speakers looked when I used REW with a crappy RadioShack SPL, which did the job. (Although thankfully my floorstanders have much better low end performance). This is the actual real world truth and unlike anything that Dirac/XMC has produced for any user on these forums with any speakers and with any room. When you look at this graph you will all just KNOW that they are probably bookshelf speakers being measured (or possibly a 100Hz crossover is in place). You similarly instinctively understand what I mean when I say my floorstanders have better low end. You can picture the graph of this in your head because you KNOW what it looks like which is different than any of the graphs from Dirac/XMC.
Our desire for the Emo to be everything we want it to be is blinding us to accept what the Dirac/XMC is telling us which cannot possibly be correct. It defies all logic. It is not an in-room response which is doing this. The data is baloney simply because it must be, no matter how much we want it not to be.
If our speakers were truly rolling off at 7kHz, then we would all be happy with compact cassettes and a Sony Walkman driving our amp via it's headphone out and not even need CD's 16 bit/44.1kHz capability yet alone hi-rez music files. 8 bit 20kHz files would be more than revealing enough and more than our speakers could resolve. |
|
klinemj
Emo VIPs
Official Emofest Scribe
Posts: 15,094
|
Post by klinemj on Jan 21, 2015 6:51:39 GMT -5
markc if you follow what Tony has been posting, it is pretty clear the calibration file is not matching the mic, and thus Dirac thinks the highs are depressed and lows are elevated. So, it is boosting the highs and cutting the lows more than it should. We will see what emo and Dirac say, but I would need to see some compelling data from them to think otherwise. Mark
|
|
|
|
Post by markc on Jan 21, 2015 7:45:29 GMT -5
markc if you follow what Tony has been posting, it is pretty clear the calibration file is not matching the mic, and thus Dirac thinks the highs are depressed and lows are elevated. So, it is boosting the highs and cutting the lows more than it should. We will see what emo and Dirac say, but I would need to see some compelling data from them to think otherwise. Mark Mark, I have been following quite closely and the measured vs expected volume differences on many peoples graphs are up to 10dB or more when measuring at 75dB target. This would mean that huge error corrections are necessary in the calibration file.
As it is a logarithmic scale 10dB error refers to a 10x power difference or 3x the amplifier voltage. That is quite an error.
It would have to be a pretty bad mic design to have this inaccuracy which is why I disbelieve that the generic emo calibration file provision, even an incorrect one at that, is the sole problem. Usually, mic calibration files have variances of 0-3dB for frequencies 100-15000Hz
Mark
|
|
|
|
Post by ansat on Jan 21, 2015 8:34:01 GMT -5
markc if you follow what Tony has been posting, it is pretty clear the calibration file is not matching the mic, and thus Dirac thinks the highs are depressed and lows are elevated. So, it is boosting the highs and cutting the lows more than it should. We will see what emo and Dirac say, but I would need to see some compelling data from them to think otherwise. Mark Mark, I have been following quite closely and the measured vs expected volume differences on many peoples graphs are up to 10dB or more when measuring at 75dB target. This would mean that huge error corrections are necessary in the calibration file.
As it is a logarithmic scale 10dB error refers to a 10x power difference or 3x the amplifier voltage. That is quite an error.
It would have to be a pretty bad mic design to have this inaccuracy which is why I disbelieve that the generic emo calibration file provision, even an incorrect one at that, is the sole problem. Usually, mic calibration files have variances of 0-3dB for frequencies 100-15000Hz
Mark
Mark, what I suspect is really going on is that the calibration used on the reference microphone is the on axis calibration and the emotiva microphone calibration was generated off that. The difference between the on axis and 90 degree calibrations is a rolloff starting around 2khz. The was a 5db difference on 2 of my cross spectrum calibrated microphones. The 6db from 10hz to 100hz is a bit puzzling. In all the testing and graphs posted and sent my way, I have run into 4 microphones that exhibit the big hump noted above 3 have already been replaced, and I am waiting on a response to confirm 1 of these fixed the issues (the other 2 issues were corrected by the microphones). For the other 2 trends I am seeing, the only real microphone variation is is between 10hz to 20khz and after my cal is applied, I would say the microphones are within tolerance to each other to 18k (which I think is plenty). For LE, I think that a corrected cal file is good enough. For the full version, my thoughts differ greatly. Tony |
|
|
|
Post by rogersch on Jan 21, 2015 8:57:43 GMT -5
Ok - So I have been seeing 3 different trends in the files posted here or the files sent to me in private. There is the everything looks like a big hump graphs. The graphs with the issues starting at 2khz - 3khz and having a slight drop-off (6db) and the 2khz - 3khz having a major drop-off around 10k (up to 12db). The big hump -- Its either the speaker or the microphone. Look up some graphs of what you speaker should look like or talk to the manufacturer of your speaker and explain what you are seeing. If you suspect that the speaker is right -- call emotiva, you likely have a bad microphone. For the other two. I have looked at a few microphone comparison files and I have come up with 2 different calibrations. One is just adding the differences between a 90* and 0* to Emotiva's existing calibration (5db correction) as well as fixing the lower end that is the same on all but one file that I reviewed. (6db starting at 100hz) My particular microphone also had an issue between 300 and 800 hz (3db) which I corrected in addition to the above corrections and microphone did not end at 5db off at 20khz I would recommend that the math cal be used as it should fit the most microphones out there. However, I have also included the calibration that is matched to my microphone. My cal filesto use the calibration - you simply need to close Dirac. replace the calibration file found at C:\Users\YOURUSERNAME\AppData\Roaming\Dirac\OEM\Emotiva\MicCalFiles (windows 8) with one of mine. As soon as these files are no longer needed (Emotiva/Dirac makes the corrections)(Or give me something to show that I am just imagining things) I will remove the files from my server and remove this post(maybe the entire thread). Tony I'll try the Math CAL on Friday morning if nobody is at home! Great work. |
|
|
|
Post by barrak on Jan 21, 2015 10:40:33 GMT -5
Markc, not that I'm disagreeing with you, especially when it comes to 5+ kHz measurements... but the graphs I posted were for a center speaker that houses dual 12" drivers in a bass reflex cabinet. I have no reason to suspect my mic down there.
If Tony is on the mark regarding the possibility of mic calibration being performed while pointing up vs pointing head-on, then this may fully explain the exaggerated early roll offs. In this case, the issue could be resolved with editing the target curve (when available) to follow the measurement past 5-7 kHz. This should be done anyway in most situations, regardless of calibration file accuracy.
|
|
KeithL
Administrator 
Posts: 10,273
|
Post by KeithL on Jan 21, 2015 13:06:38 GMT -5
Really, guys.  A speaker (or individual driver) starts to become significantly directional when the wavelength of the frequency involved becomes SMALLER than the dimension of the radiating surface. This is one way to control (or predict) dispersion in a particular direction. For a round cone or dome in a regular cabinet, the "dimension" is the diameter of the driver, so they become seriously directional on all axes when the wavelength at whatever frequency you're describing becomes shorter than the diameter of the driver. The wavelength at 2 kHz is about 6", and becomes shorter as you go up in frequency, which means that it is quite normal (and virtually unavoidable) for an 8" cone to become somewhat directional at any frequency above 1 kHz or 2 kHz, and to become more directional as the frequency goes higher. Note that, in some circumstances, for a center channel speaker, this could be considered desirable. Because the speaker is directional at even midrange frequencies, most of the energy goes forward towards the listener, so less ends up arriving at the listener later after being reflected off the floor, ceiling, and walls. (In other words, the larger driver has "controlled vertical and horizontal dispersion at voice frequencies", which could serve to improve voice clarity and imaging, especially in a very live room.) The ability of a speaker to make low bass is limited by both its frequency response and its displacement. Even a tiny speaker can be designed to produce frequencies down to 20 Hz - just not very loudly. (Many headphones can deliver sound flat to 20 Hz with a 1" or 2" speaker - but only into a "room" the size of your ear canal.) Most normal speakers deliberately trade off the ability to produce very low frequencies in return for the ability to produce reasonable sound levels. The reason you need a subwoofer is to produce enough displacement of air to produce low notes at a satisfying volume level. The ability of a speaker to make high frequencies (and without distorting) is determined by a large number of factors, but it's not at all unusual for an 8" driver to start rolling off anywhere above a few kHz, and to be able to make some usable output down to 20 Hz, and it would be downright odd if an 8" driver DIDN'T become very directional above a few kHz. And any speaker that uses multiple drivers to produce an overlapping range of frequencies will also almost certainly be subject to assorted comb filter and interference effects between those drivers as well. Every room and every speaker is different. We've compared the measurements taken with our microphones (and our correction curve) to ones taken IN THE SAME ROOM, UNDER THE SAME CONDITIONS with a "serious" calibrated measurement microphone - and they DO agree within a very small margin of error - and the microphones we've spot checked are ALL very consistent in their responses. Our microphone is quite flat AT 90 DEGREES OFF AXIS - AS THE INSTRUCTIONS SPECIFY with our correction curve. I don't know the details of exactly how Dirac evaluates its data, or how it weights the direct and reflected energy at various frequencies when calculating its corrections, so I don't at all assume that it will produce the same results as REW - especially under different conditions and with different rooms. (In fact, since Dirac is specifically correcting things that REW doesn't even CLAIM to fix, I would hope the results would be rather different.) The Full version of Dirac will be available shortly, and then you'll be able to use a different microphone if you like, and create your own Target Curves.... By all means, if you LIKE the results you get with REW better than the ones you get with Dirac, then use REW... we don't mind a bit...
HOWEVER, complaints about how "the measurements must be wrong because I just know my speaker couldn't measure like that" and "we shouldn't believe measurements like that" are just plain tedious. And expecting someone who hasn't heard your system or your room to interpret your graphs in detail is an exercise in futility. I'll also admit I'm starting to get a bit cheesed off with the incessant banter about how "the measurements can't be right" and "the graphs can't look like that". If the onscreen graphs annoy you, then just close your eyes! The whole point of Dirac is to adjust the way your system SOUNDS; not to adjust graph so it looks prettier. When the Full version of Dirac comes out (which will be quite soon now), you'll be able to create and adjust your own Target Curve, which will allow you to adjust the way your results sound, at which point you can use the graphs as A GUIDELINE so you can see what you want to adjust and how... Please understand that I'm all for asking someone with more knowledge and experience if they recognize specific issues on a graph - like room modes or crossover anomalies - but trying to figure out "if your graph is accurate" by comparing it to SOMEONE ELSE'S MEASUREMENTS is just plain silly. In the mean time, if your real goal in all this is to produce a nice looking graph, and you're too busy doing that to listen to any music, I can recommend a few nice programs that will allow you to do that very easily, while entirely avoiding all that annoying listening and measuring... personally I prefer Adobe Photoshop.   Single point or 9 point measurement? I think the issue is that none of us should believe that a centre speaker (limited volume cabinet, possibly smaller drivers, internal crossovers optimised for dialogue clarity etc) could possibly have such AMAZING low end output (from 20Hz up) and yet a massive roll off starting at 5kHz.
Speakers are just not made this way and that is why we use subs and why a sub woofer is usually run with a crossover between 50-120Hz to compensate for the inability of standard speakers to reproduce these frequencies and why we only need one tweeter irrespective of the number of mid/low range drivers in the cabinet - they can just produce the correct high frequencies easily.
We are not measuring what we think we are measuring and it is a microphone problem or a calibration file problem or possibly something going on in Windows / Dirac. Worse, we are believing what we are seeing on screen when it cannot possibly be correct and debating plausible workarounds to modify it.
Unless the calibration file for the Emo mic has some incorrect and major positive corrections at the low end and similarly aggressive negative corrections for the high frequencies, then the mic itself is not measuring correctly.
I have only ever used an analogue mic into soundcard for REW rather than the USB interface. Could this be somewhere that a problem lies rather than the EMO mic?
Is it the test tone? Is it incorrect or not what Dirac is expecting?
Here is a picture from the UMIK page as their example of how a speaker should measure in Dirac, which shows what we would all expect a full range speaker to look like, not like ANY of the screenshots of XMC users. The trace climbs up to 50/60Hx and is pretty steady until a roll off starting 17kHz.
www.minidsp.com/images/appnotes/umik-1-dirac/umik-1-dirac-measured.png
~Here is a picture from the Dirac website. It also show us what we know a speaker response should look like. Acceptable performance from 50-20kHz with only the expected and usual variations across the trace due to nulls/voids/reflections etc
Here is someone's REW measurement using a UMIK mic from the REW forums. This is the same pattern as how my speakers looked when I used REW with a crappy RadioShack SPL, which did the job. (Although thankfully my floorstanders have much better low end performance). This is the actual real world truth and unlike anything that Dirac/XMC has produced for any user on these forums with any speakers and with any room. When you look at this graph you will all just KNOW that they are probably bookshelf speakers being measured (or possibly a 100Hz crossover is in place). You similarly instinctively understand what I mean when I say my floorstanders have better low end. You can picture the graph of this in your head because you KNOW what it looks like which is different than any of the graphs from Dirac/XMC.
Our desire for the Emo to be everything we want it to be is blinding us to accept what the Dirac/XMC is telling us which cannot possibly be correct. It defies all logic. It is not an in-room response which is doing this. The data is baloney simply because it must be, no matter how much we want it not to be.
If our speakers were truly rolling off at 7kHz, then we would all be happy with compact cassettes and a Sony Walkman driving our amp via it's headphone out and not even need CD's 16 bit/44.1kHz capability yet alone hi-rez music files. 8 bit 20kHz files would be more than revealing enough and more than our speakers could resolve. |
|
|
|
Post by Priapulus on Jan 21, 2015 13:45:17 GMT -5
Really, guys. HOWEVER, complaints about how "the measurements must be wrong because I just know my speaker couldn't measure like that" and "we shouldn't believe measurements like that" are just plain tedious. And expecting someone who hasn't heard your system or your room to interpret your graphs in detail is an exercise in futility. I'll also admit I'm starting to get a bit cheesed off with the incessant banter about how "the measurements can't be right" and "the graphs can't look like that".
Well, the graphs don't mean much to me, but it cheeses me off, when EMO's chief engineer dismisses me, because I'm upset that when I turn on Dirac and play organ music, the pedal disappears, and the upper work screams. I don't know much about mikes and speakers, but I do know what organs sound like.
Without Dirac I hear pedal notes and balanced trebles. With Dirac the pedal disappears and the mixtures scream.
Do I just accept pedalless music? Really? Is it really my fault? Maybe I've got a bad mike, maybe you provided a bad calibration, maybe Dirac phucked up. I'm appalled you hear a chorus of concerns, and don't think that it's your problem. And appalled at your criticism of the people trying to help you sort your problem.
Sincerely /b
|
|
|
|
Post by rogersch on Jan 21, 2015 13:54:57 GMT -5
Ok - So I have been seeing 3 different trends in the files posted here or the files sent to me in private. There is the everything looks like a big hump graphs. The graphs with the issues starting at 2khz - 3khz and having a slight drop-off (6db) and the 2khz - 3khz having a major drop-off around 10k (up to 12db). The big hump -- Its either the speaker or the microphone. Look up some graphs of what you speaker should look like or talk to the manufacturer of your speaker and explain what you are seeing. If you suspect that the speaker is right -- call emotiva, you likely have a bad microphone. For the other two. I have looked at a few microphone comparison files and I have come up with 2 different calibrations. One is just adding the differences between a 90* and 0* to Emotiva's existing calibration (5db correction) as well as fixing the lower end that is the same on all but one file that I reviewed. (6db starting at 100hz) My particular microphone also had an issue between 300 and 800 hz (3db) which I corrected in addition to the above corrections and microphone did not end at 5db off at 20khz I would recommend that the math cal be used as it should fit the most microphones out there. However, I have also included the calibration that is matched to my microphone. My cal filesto use the calibration - you simply need to close Dirac. replace the calibration file found at C:\Users\YOURUSERNAME\AppData\Roaming\Dirac\OEM\Emotiva\MicCalFiles (windows 8) with one of mine. As soon as these files are no longer needed (Emotiva/Dirac makes the corrections)(Or give me something to show that I am just imagining things) I will remove the files from my server and remove this post(maybe the entire thread). Tony Hello Ansat, I did an comparison in excel of the my original CAL file (orange line) and the CAL file provided by you (Math version, blue line in the graph). See graphic below:  I can understand that "math" CAL file will lead to more Bass and less Treble but I just don't understand the bump in the range from 485 Hz to around 800 Hz. Are you able to explain this? |
|
