or Connect
AVS › AVS Forum › Audio › DIY Speakers and Subs › Dayton OmniMic Precision Measurement System
New Posts  All Forums:Forum Nav:

Dayton OmniMic Precision Measurement System - Page 38

post #1111 of 2086
Smoothing is fine to an extent. I think 1/6th octave smoothing is about what our ears do on their own anyways.
post #1112 of 2086
Quote:
Originally Posted by MKtheater View Post

How does one fix the upper bass in that graph with no smoothing? There seems too many peaks and nulls, positioning of the subs?

Why do people run Audyssey and report at how much better it sounds? Placement, phase, delays, what type of mains, X-over tweaks... there are lots of ways to smooth a FR and I'm a bit surprised to hear you're not aware of them?

Quote:
Oh yeah to the people that questioned if it was smoothed or not, look at bosso's unsmoothed graph, you clearly see how the 30hz and below is less peaky than upper bass.

You're missing the point. Compare my unsmoothed REW graph to your supposedly unsmoothed graph. When I crop my graph at 30 Hz and stretch your graph to fit my graphs scale:



Maybe I'm the only person who can see the resolution difference (jagged limited data point trace vs normal hi-rez trace)??

Bosso
post #1113 of 2086
Quote:
Originally Posted by LTD02 View Post

ears and eyes don't hear the same.

Are you serious?

Of course they do, or are measurements and tweaks a useless waste of time and money?

Ignorance is bliss? Yeah, I'll pass.

Bosso
post #1114 of 2086
Quote:
Originally Posted by GPBURNS View Post

not sure why you don't get it unless we are talking different topics - it happens - no worries

Quote:
Originally Posted by bossobass View Post

All I'm saying is that the graph you posted is 1/6 octave.

And it seems to be spreading.
post #1115 of 2086
Quote:
Originally Posted by ironhead1230 View Post

I started looking into about adding a calibration file and this is what I found. It looks like it would be a little bit of work but can be done.

http://tech.groups.yahoo.com/group/S...e&var=1&tidx=1

http://www.qsl.net/dl4yhf/speclab/fi...htm#fft_filter

So we don't take this thread too far off topic, I started another thread about Spectrum labs captures to get these questions answered. I want to get a full how to written with a common settings file for anyone to use.
http://www.avsforum.com/avs-vb/showthread.php?t=1374817

-Mike

Nice find ... thx! Now if you could just find some more time in the day for me ...
post #1116 of 2086
Quote:
Originally Posted by Gary J View Post

And it seems to be spreading.

Delusion is a funny phenomenon.

But seriously, if we're going to start posting low-rez measurements on cropped and grossly stretched graphs with tons of data missing, then I'd prefer if we went back a decade in time. Back then, I used the RS meter, a test tone CD (which I still have) and a standard 'C' weight RS meter correction file. I plotted each data point on a hand made graph and simply drew a straight line between each data point. The result was exactly like the graph posted above with one exception... Even a decade ago I used 1/12 octave test tones.

For those interested (and most of you know this already, but some may not grasp the significance), when you have single data points, the trace plots a straight line between those points. The lower the resolution (lower # of fractions of an octave), the higher the chance that there is something significant occurring BETWEEN the data points that will simply be ignored by the measurement.

A hi-rez measurement plots a 'curve' that connects infinitely more data points, after which measurement "smoothing" may be applied to the trace, as shown in the 3 graphs I posted earlier. Smoothing has its place in tweaking a system, but it serves as a detriment when tweaking in-room response below 100 Hz.

In the same vein, and this is something I dislike about the GUI of Omnimic, squashing the vertical scale and stretching the horizontal scale of a graph so that the trace is radically out of scale with what we've all been used to seeing from Ed Mullen, AVTalk, Ilkka and REW has the same effect as lowering resolution to most casual observers.

FR, SL, decay and other graphs are extremely useful and have shaped the state of the art far more than any other single advance, but in the end you have to CREATE, then READ the graph accurately.

I dislike the current trend of cheating measurements, downplaying non-linear behavior, fudging data into phake specs, throwing 2M RMS numbers out the window in favor of 1M Peak numbers to grab bragging rights, using modeling software to pretty up a less than optimal trace, claiming theoretical output at 5 Hz when using an amp that gives nothing at 5 Hz, etc.

Elsewhere, I couldn't care less, but here is where we've gone to great lengths to get to the bottom of taming the beast. It would be idiotic to make a u-turn and go Best Buy after all of that effort.

Bosso
post #1117 of 2086
It seems like between number of data points, resolution of test signal, smoothing of measurement data, equipment and technique two people doing basically the same thing can come up with something looking really smooth or really jagged.
post #1118 of 2086
Quote:
Originally Posted by bossobass View Post

Why do people run Audyssey and report at how much better it sounds? Placement, phase, delays, what type of mains, X-over tweaks... there are lots of ways to smooth a FR and I'm a bit surprised to hear you're not aware of them?



You're missing the point. Compare my unsmoothed REW graph to your supposedly unsmoothed graph. When I crop my graph at 30 Hz and stretch your graph to fit my graphs scale:



Maybe I'm the only person who can see the resolution difference (jagged limited data point trace vs normal hi-rez trace)??

Bosso

I understand that it has less resolution but I was just saying on omnimic I selected no smoothing. I could have made this a straight line with 1/6th. Either way apparently I was lucky to get this graph because I amm having a hard time getting a graph now. I am pulling out my calibrated to 5hz behringer MIC(I hope) and pre mobile and run some REW graphs.
post #1119 of 2086
MK,

What issue(s) are you having getting readings from Omni?

James
post #1120 of 2086
Quote:
Originally Posted by exojam View Post

MK,

What issue(s) are you having getting readings from Omni?

James

It is showing that I have a 50hz and 100hz peak of 30db's with track 2 and sine waves. It then shows on distortion or bass decay that I have a 10-12 db peak at 50hz only. I don't know what to believe. Also, my graph keeps changing all the time on response and I don;t know which one to capture. Everyone tells me their response stays the same or hardly moves, mine changes too much. Something has to be wrong.
post #1121 of 2086
For the sine wave - track 2 and distortion - track 3, make sure the smoothing is the same. If not, one may bounce around a bit more than the other.

One thing you could try is to run a trace, save the FRD file, than repeat this 3or 4 times (you shold be able to do this before the track ends). Stop measuring. Then go back in and load those saved FRD files from the Curves - Add, then check how different they are. If nothing is moved like the mic, volume, etc they should be pretty close.

Just as a FYI, when I ran the distortion graphs, it would usually take a few sweeps to see changes on the screen.

James
post #1122 of 2086
how much lower res. are the graphs for OM than REW?
post #1123 of 2086
When Omni-Mic measurements are taken can you specify what frequency range is being measured or does it always capture the same measurement of the same sweep duration? I am wondering whether the basic resolution of the measurements is fixed and if what is shown in MK's 5-30Hz graph is the maximum number of data points.

What file types can be exported out of Omni-Mic? Can a .txt file be exported? I'd be interested to see this for a measurement with no smoothing.

REW and other audio programs resolution over a certain frequency range can be increased or decreased by adjusting the measurement signal duration and/or restricting the bandwidth covered. For example REW can provide 87 data points below 30Hz if using the longest duration sweep and restricting the measurement to 100hz.
post #1124 of 2086
Quote:
Originally Posted by Ricci View Post

When Omni-Mic measurements are taken can you specify what frequency range is being measured or does it always capture the same measurement of the same sweep duration? I am wondering whether the basic resolution of the measurements is fixed and if what is shown in MK's 5-30Hz graph is the maximum number of data points.

Sweep content is fixed since it's played from a CD. Chart resolution is adjustable from 3ppo to 96ppo. Frequency upper and lower limits, as well as vertical scale are adjustable.
post #1125 of 2086
The finest resolution in Hz for OmniMic frequency response measurements is 48000/32768=1.46Hz, the sweeps are spaced 683msec (which is also the inverse of that number). The sample rate divided by the sample array length. Are you saying that 1.5Hz resolution isn't enough (three cycles every 2 seconds)? What features in your room do you expect to cause that kind of narrow peak or null? -- that's pretty much down in the pressurization region of frequencies except maybe in a large cave or a shopping mall.

The other "Nth Octave" resolutions are smoothed from that inherent resolution, but it can't be made from thin air to extend it further.

I can't say whether the red curve shown is higher res or not, it could also just be spline curve fits between the points to make it look smooth (not difficult but not really meaningful either), or it could be from a much, much longer time acquisition. But take the time duration of each data collection that made it and invert it (Hz from 1/sec) and that's the best resolution it could be, do the math.
post #1126 of 2086
Quote:
Originally Posted by RUR View Post

Sweep content is fixed since it's played from a CD. Chart resolution is adjustable from 3ppo to 96ppo. Frequency upper and lower limits, as well as vertical scale are adjustable.

is 96ppo the out of the box setting or do you have to bump it up?
post #1127 of 2086
Ricci,

I am not sure if this answers your question concerning the data points or not but here it goes.

Here is a section of a 1-6 smoothing FDR I have, I am only copying to the end of 13HZ:

Frequency[Hz], dBSPL
11.86765 69.1
12.04027 69.09
12.2154 69.07
12.39308 69.03
12.57334 68.98
12.75622 68.89
12.94176 68.77
13.13001 68.58
13.32099 68.34
13.51474 68.1
13.71132 67.89
13.91076 67.69

Here is another section of a trace taken at the same time as the one above but using 1-24th smoothing, again only going to the end of the 13HZ:

Frequency[Hz], dBSPL
11.97524 69.28
12.01856 69.08
12.06202 68.9
12.10565 68.77
12.14943 68.68
12.19337 68.61
12.23747 68.56
12.28173 68.53
12.32615 68.5
12.37073 68.48
12.41547 68.47
12.46037 68.46
12.50543 68.46
12.55066 68.45
12.59605 68.45
12.64161 68.45
12.68733 68.45
12.73322 68.44
12.77927 68.44
12.82549 68.44
12.87187 68.44
12.91842 68.44
12.96515 68.44
13.01204 68.44
13.0591 68.44
13.10633 68.43
13.15373 68.43
13.2013 68.42
13.24905 68.41
13.29696 68.4
13.34505 68.38
13.39332 68.36
13.44176 68.33
13.49037 68.28
13.53916 68.22
13.58813 68.17
13.63727 68.13
13.68659 68.1
13.73609 68.09
13.78577 68.07
13.83563 68.06
13.88567 68.06
13.93589 68.05
13.98629 68.05

I only went to the end of 13HZ for space purposes in the thread.

James
LL
post #1128 of 2086
Quote:
Originally Posted by beastaudio View Post

is 96ppo the out of the box setting or do you have to bump it up?

If you're asking the default setting, I think it's 6ppo.
post #1129 of 2086
Quote:
Originally Posted by bwaslo View Post

The finest resolution in Hz for OmniMic frequency response measurements is 48000/32768=1.46Hz, the sweeps are spaced 683msec (which is also the inverse of that number). The sample rate divided by the sample array length. Are you saying that 1.5Hz resolution isn't enough (three cycles every 2 seconds)? The other "Nth Octave" resolutions are smoothed from that inherent resolution, but it can't be made from thin air to extend it further.

Thanks Bwaslo. (You too Exojam). I don't have an Omni-Mic myself. That answers my question.

There is plenty of resolution available it just needs to be utilized when taking the measurements.

BTW there isn't anything wrong with smoothing but it really should be used sparingly to clean up noisey or rough data or to make a collection of data easier to decipher. I like 1/12th octave for this myself as it cleans things up but still leaves plenty of fine detail and largely leaves large features untouched. I don't usually use any more smoothing than that for <200Hz.
post #1130 of 2086
Quote:
Originally Posted by bwaslo View Post

The finest resolution in Hz for OmniMic frequency response measurements is 48000/32768=1.46Hz, the sweeps are spaced 683msec (which is also the inverse of that number). The sample rate divided by the sample array length. Are you saying that 1.5Hz resolution isn't enough (three cycles every 2 seconds)?

Is that finest resolution reflected in the posted OM graph?

Bosso
post #1131 of 2086
MK,

Did you by chance save that graph in question as a frd file? Thanks.

James
post #1132 of 2086
Quote:
Is that finest resolution reflected in the posted OM graph?

Depends on the which graphs. It can't be better (less) than 1.5Hz, but it can be coarser via smoothing. 1.5Hz is only about a tenth octave at 20Hz, but it is near 1/70th octave at 150Hz, and very very small (in octaves) by 1kHz or more.

Points can be remapped to finer frequency point pitch (could make the graphs show thousands of points below 10Hz), but the actual resolution included in that data is limited to "how long you look at it". If you want resolution to 1/100th of a Hz, then you have to look at s signal at the very least 100seconds long.

FRD data points reported is not the same as resolution. Similar to that you can look at 480i video upscaled to 1080p, but it is still 480i that you are really looking at....
post #1133 of 2086
Quote:
Originally Posted by exojam View Post

MK,

Did you by chance save that graph in question as a frd file? Thanks.

James

I don't know, I just hit the snapshot so it is most likely just a Jpeg.
post #1134 of 2086
Quote:
Originally Posted by bwaslo View Post

FRD data points reported is not the same as resolution.

Similar to that you can look at 480i video upscaled to 1080p, but it is still 480i that you are really looking at....

Why isn't it is my question?

If it was collected there why is it not incorporated into the resolution? If it was not collected then where does this data come from and what is the point of it?
post #1135 of 2086
Quote:
If it was collected there why is it not incorporated into the resolution?

Because when smoothing is done and data is converted from "A points per Hz" to "B points per Octave" frequency data points get added or removed. Otherwise you would have 1600 points to plot in the octave from 10kHz to 20kHz and just a few from 10Hz to 20Hz. Sound frequency is perceived in octaves, but FFTs give it to you at Hertz points.

Quote:
If it was not collected then where does this data come from and what is the point of it?

It is generally formed by making a curve of the obtained data points and mapping to different points (perhaps more or perhaps fewer) on that curve. Think of interpolation, it is like that only done more like with a french curve than with a ruler. The point (no pun intended) is that you are interested in the shape a response not just dots at discrete points.

If you only looked at 480 pitch on your 1080 TV set, you'd have an awful lot of unused pixels.
post #1136 of 2086
This thread is really picking up steam. Glad I got the OmniMic now.
post #1137 of 2086
Quote:
Originally Posted by Scott Simonian View Post

This thread is really picking up steam. Glad I got the OmniMic now.

Me too I just really want it NOW so i can go play some as suggestions are being thrown out!!
post #1138 of 2086
Quote:
Originally Posted by bossobass View Post



Quote:
Originally Posted by bwaslo View Post

Depends on the which graphs. It can't be better (less) than 1.5Hz, but it can be coarser via smoothing. 1.5Hz is only about a tenth octave at 20Hz, but it is near 1/70th octave at 150Hz, and very very small (in octaves) by 1kHz or more.

Points can be remapped to finer frequency point pitch (could make the graphs show thousands of points below 10Hz), but the actual resolution included in that data is limited to "how long you look at it". If you want resolution to 1/100th of a Hz, then you have to look at s signal at the very least 100seconds long.

FRD data points reported is not the same as resolution. Similar to that you can look at 480i video upscaled to 1080p, but it is still 480i that you are really looking at....

I understand the points you're making well enough, but the question isn't being answered. Of course it depends on what graph, that's why I was referring to the only graph I've been talking about in this thread.

You commented on the red trace, so I assumed you couldn't miss the black trace, which is the graph I'm referring to.

Yes, data points reported is the same as resolution in context with my Q. I'm speaking about reading a graph, which is all we can do when one is posted. If a graph is smoothed to 1/6 octave, it will simply have less resolution than the no smoothing graph of the same measurement.

MKT has said that he has selected 'no smoothing' for that graph.

If the graph indeed shows the result of selecting the no smoothing option, is that trace typical of the OM result?

I don't know how else to ask the Q.

Bosso
post #1139 of 2086
Quote:
Originally Posted by bwaslo View Post

Because when smoothing is done and data is converted from "A points per Hz" to "B points per Octave" frequency data points get added or removed. Otherwise you would have 1600 points to plot in the octave from 10kHz to 20kHz and just a few from 10Hz to 20Hz. Sound frequency is perceived in octaves, but FFTs give it to you at Hertz points.


It is generally formed by making a curve of the obtained data points and mapping to different points (perhaps more or perhaps fewer) on that curve. Think of interpolation, it is like that only done more like with a french curve than with a ruler. The point (no pun intended) is that you are interested in the shape a response not just dots at discrete points.

If you only looked at 480 pitch on your 1080 TV set, you'd have an awful lot of unused pixels.

Right. I gotcha there.

You mentioned 1.46Hz but the snippet of .frd file that Exojam posted seems to contain a lot tighter packet of information than this. Doesn't the lowest octave resolution determine the maximum per octave resolution for the entire measurement as displayed? I'm no software guy, I'm just trying to understand this a little better and get a handle on the OM's resolution below 30Hz.

MK's measurement seems to indicate a very coarse resolution in that range.

Edit: And +1 on Bosso's post above.
post #1140 of 2086
Quote:
Originally Posted by bossobass View Post

I understand the points you're making well enough, but the question isn't being answered. Of course it depends on what graph, that's why I was referring to the only graph I've been talking about in this thread.

You commented on the red trace, so I assumed you couldn't miss the black trace, which is the graph I'm referring to.

Yes, data points reported is the same as resolution in context with my Q. I'm speaking about reading a graph, which is all we can do when one is posted. If a graph is smoothed to 1/6 octave, it will simply have less resolution than the no smoothing graph of the same measurement.

MKT has said that he has selected 'no smoothing' for that graph.

If the graph indeed shows the result of selecting the no smoothing option, is that trace typical of the OM result?

I don't know how else to ask the Q.

Bosso

I might be oversimplifying this, but doesn't the granularity of the black trace represent the 1.5hz resolution described within the lower octaves? Visually, the number of distinct points appear to align with the stated resolution in the 10-30 hz graphs.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: DIY Speakers and Subs
AVS › AVS Forum › Audio › DIY Speakers and Subs › Dayton OmniMic Precision Measurement System