Dayton OmniMic Precision Measurement System - Page 7

^^^^

Mmmmmmmmmm,..Vegas, I like Vegas more than HT and subs!

Heheh. I'm going for CES 2011 on the 6th. Should be fun.

I have one of those 720p HD pocket cams so I'll try and take some vids of the show. I've been going for a long time so for me, it's a 'been there, done that (seen that)' kind of thing but I'll try to capture some goodies if there is any. I have always looked for cool speaker systems but each year the outtings get lamer and lamer. My experience at the Earthquake Sound booth last year, for instance. I don't think I'll buy another regular store speaker ever again.

despite my challenges, i am in the "accurate" reproduction is best camp.

...just trying to get folks to stretch their thinking in order to see if there is something more and/or more fully define what is "accurate".

most of these spectrumlab plots are incomplete in my estimation, as they don't fully consider the amplitudes. spectrumlab is averaging. show me the source vs. reproduced waveform, as in audacity.

it is freeware, so no ante to play.

Disagree.

The original waveform is ones and zeroes. The digital-->analog conversion, which is then preamplified-->delayed-->EQ'd-->amplified and sent to the subwoofer(s) must be recorded-->preamplified and converted back to digits.

Only then does your preferred software matter, and that's a lot of hurdle before you get to that discussion.



SL compares its visual depiction of the digits to the digitized analog version offered by your subwoofer system regarding:

Harmonic distortions
Response magnitude distortions
Bandwidth distortions

I find this comparison vastly superior to dragging the system out to some parking lot and torturing it with sine waves.

Keith Yates is the only guy to ever have done this comparison, but he chose to do so outdoors. What happens in the 1st 2 octaves with any given sub in a room will happen with that sub in any room and the comparison of the SLs indoors vs outdoors would have been far more interesting to me than the rest of the sine sweep/burp baloney.

As for going into further detail than that, have at it and I'll let you know what I think.

Bosso

No, I don't.


Calibrated ECM-8000 running off an M-Audio FireWire Solo, though admittedly only in the midrange. (I still use a Velodyne SMS-1 with their MIC-5 spatial averaging kit to measure bass response in my room, and provide global EQ. I just find it quicker and easier than other solutions.)



What I'm saying is that for most of what we need to do at home the tolerance is fine. Now, should one develop a commercial product using that kind of quick-n-dirty measurement? No, especially given that small placement differences will likely have greater effects. But to figure out room problems, help with speaker placement, and even DIY crossover design work (given that the tolerances generally manifest themselves at the ends of the response, not in the middle)....it's easy to overthink things when there's a simple solution.


Agreed.


Thanks for the comparo.

Big +1.

I am interested in this and will probably pop for one.

Sooner would be better then later. I assumed this is appropriate for full frequency measurements. I have not had time to read all the documentation for it.

i set up an experiment to test my hypothesis that spectrum lab is averaging.

i created a 30hz signal of amplitude 1.0, 3 cycles long.

next i created a 30hz signal of amplitude 0.5, 18 cycles long.

the two are separated by two seconds of silence and then i played them in a loop ( and the play button in audacity). then i went over to spectrum lab and observed how the two were being recorded graphically.

here is the audacity plot:



here is the spectrum lab plot:



it can easily be seen that spectrumlab is differentiating between the two, but guess what, the brightest pink is *NOT* coming from the signal with the highest amplitude. it is coming from the signal with 1/2 the amplitude, but 6 times the cycles.

this is what i was trying to get at...spectrum lab is averaging. averaging is no good for measuring transients.

send me your email in pm and i will be happy to send you the audacity file so that you can test it out for yourself bosso.

You are correct that Spectrum Lab misses some transients. It isn't averaging though.

The program works by taking snapshots of the waveform in time, and displaying one pixel line per snapshot, if you have it setup right. This means the absolute peak can be missed due to timing. It also means the waterfall graph will emphasis high concentrations of signals to the eye, but not necessarily the peak signals or frequencies. This is why proper setup is needed.

SpecLab can show you general frequency makeup, and amplitude distribution, but that is it.

EDIT: That said... are you sure nothing is clipping there? Something seems amiss, as there are a lot of extra stray harmonics showing.

soho, what option would one use in order to turn up the resolution so that the output would not appear as "averaging".

That sounds user-ugly.

Any reason to think OmniMic will have the same pitfall?

"EDIT: That said... are you sure nothing is clipping there? Something seems amiss, as there are a lot of extra stray harmonics showing."

i knew that you were going to ask that, but i figured i'd leave the mess in, as that seems to be what most folks do in these spectrumlab plots anyways. the same effect is present even when i turn the amplitudes way down. you were on my case about that before and i thank you.

argh. somehow, now my spectrumlab is all screwed up. wtf. it appears to have defaulted back to all its standard settings. maybe it was the shockwave from santa's sleigh, but it appears that i'm going to have to reconfigure everything. great present. :-( when it rains it pours. bosso, did you hack into my computer: ;-)

ok, i'm not making this up. i may have demons in my computer.

during the last response, to soho, i refired spectrum lab and the audacity file. all the colors were botched, the frequency was botched, essentially everything was screwed up. the program was behaving like something else.

as soon as i sent the note, i clicked back to spectrum lab and all was how i set it up. i've never seen this kind of behavior. it was like an entirely different program was running.

very, very weird. first time that i have ever seen anything like that.

@ LTD02.... maybe what you need is a Computer Exorcist...

Video of exorcism here:

http://video.google.com/videoplay?do...6380216775608#

Even funnier video here : http://www.youtube.com/watch?v=no3rvduw1d8


One such practicing holy man / excorcist.... :P

http://www.computerexorcist.com/

OP: Sorry for the thread jack, I couldn't help myself

lol. i can't blow up my computer mr. hurd--it's not like the fax in office space.

http://www.romston.com/2009/09/05/mo...achine-moment/

i'm not going gangter on it. the computer exorcist bit has me laughing so hard that i had to call the number. i just got plopped into voice mail. shucks, that would have been hilarious. i was planning to play along and see where it went.

the spectrum lab thing was very real and very weird though.

This is not good,...I've got an IB install in the balance!

I'll call.

I like to use spectrum lab to set up my gains and look at the output signal of electronics. LTD02 are you saying that you can't get it to show a pure sine wave input without harmonics? They should easily be way way down in level unless something is clipping in your chain. I also calibrate its spl levels shown. I tend to set it up for a faster response time but with less FR resolution. You tend to trade one for the other. You can miss the peaks but capture a really detailed frequency range or you can get more of the peaks but deal with larger chunks of frequencies. I use a much faster scroll speed than most do as well. Maybe Soho knows how to have the cake and eat it too?

LTD,

Looking, for example, at Ilkka's SW tests, no one ever said they were definitive. The value is in comparing the tested subs results against one another for a relative comparison.

There are accurate ways to measure transients, but I never claimed SL was the tool to use. I'm interested only in gauging the results of an actual subwoofer system in an actual room.

I believe that, given accurate audio measurement hardware, SL offers the best overall tool to do that vs what we've been used to seeing from the sine wave torture testing in a parking lot.

Despite the excellent work of Mullen, CJ and Ilk, they leave out the 1st 2 octaves of the LFE BW, they don't (can't) factor in boundary gain or pressure vessel gain (the combined effect being referred to as Room Gain) and they can't allow one to "see" what he's hearing vs what's encoded on the disc.

Over the years, literally hundreds of people have reported "useable" output from their ported subs as far as an octave below tune. Knowing the dynamic tracking abilities of a subwoofer can't dispute or verify such a claim, but SL can... vividly.

If you look at any 20Hz tuned ported sub with HP protection, it's down -30 to -40dB at 10Hz with triple-digit THD. I know of no room that can reconcile that, much less render the output at 10Hz "useable", considering the huge drop in output and the amount of harmonic distortion that will result.

Yet, time and again, we see the in-room FR graphs and read the claims.

SL would lay those claims and inaccurate FRs to rest in a single graph.

All I'm saying is that SL is the best tool I know of to relatively compare actual performance vs the digital feed.

This is why I'm happy to see that PE is offering a much better (on paper) alternative to the RS meter-as-mic graphs that dominate the scene, and am looking forward to FOH's and Warp's reports and comments.

Bosso.

The waterfall can miss "amplitude peaks". In theory the peak holding spectrum graph is more accurate.

------------------

"Peak-holding spectrum graph

If this option is set, the spectrum graph shows an additional curve with the peak value of the previous N seconds. The number of seconds can be set between 0.5 and 60 . This function is good for reading the amplitudes of "short tone bursts" or similar, or to display the result of a frequency sweep. For this reason, the peak values can be cleared and "frozen" via interpreter command too.

The colour of the peak curve in the spectrum graph can be modified here.
Note: The actual peak detection is done more frequently than the display update ! Some "peaks" may have such a short duration, that you won't recognize them in the momentary spectrum graph, but only in this peak indicator."

Check out the debugging window. If you scroll to fast the program tells you that you are scrolling too fast.

Can be found by going to View/Windows, Debugging Window, MISC tab and error history tab. Look at the other tabs while you are at it.

A quick note. This is a spectrogram of the opening salvo scene of Master and Commander. The left side is LFE alone, the right side is Rm C and L combined. The upper spectrogram is the PEAK HOLD (green line).









When I are interested in the distribution of amplitude peaks, then I look at the PEAK HOLD spectrogram by itself. Same opening salvo scene that shows amplitude peaks. Infra below 20 Hz is much lower in level than 30 to 200 Hz. Note that the peaks are rolling off starting at around 30 - 40 Hz.

Have a happy holiday guys!

That's what you get with a test signal like this. The sudden turn on and turn off of the signal generates HF stuff, mostly odd-order harmonics. It's kinda like a square wave being made of the fundamental plus the harmonics.



To avoid that, you either have to let it run long enough that the harmonics don't matter or you have to use a Linkwitz-style tone burst that ramps up and ramps down gently. FWIW, the latest version of ARTA does the Linkwitz tone bursts.

dennis, your are right. when i run a very long signal, it is just the signal, but when it starts and stops, there is lots of broadband noise. since the example that i provided has relatively quick starts and stops, there is lots of broadband noise. you nailed it.

dave, i think we are on the same page now. spectrum lab cannot be used to measure amplitude peaks accurately in the color plots, but it probably can be used to show the capabilities of one subwoofer vs. another with respect to some audio track.

holiday cheers.

We have been though this exact conversation before.

What is shown is not the normal start up/short wave response. You are overdriving an input or output somewhere in the signal chain.

A short burst (50Hz-5cycles) looks like this:

We have covered this as well. it is also why all of my test sugnals have employed this technique over the years.

You get this:


Instead of this: