So ... we've got a FR/distortion rig... now what?

[brianpowers27]

The measurement rig is mostly calibrated and good to go. I have got some beer and some test tones via Arta. Tonight may be the first time that I have been confident with the measurements I have taken.

Preface
I took a slew of measurements on 4 different drivers. I messed around with multi-tone sine waves and found some of the results to be revealing. Like any scientist, I am left with more questions than answers. (We shall have to sidebar the IMD testing discussion.) Of particular interest for tonights discussion are the distortion plots and waterfall/burst plots. This will help us to quantify the particular character of the drivers under test.

Goals
--Familiarize myself with testing rig
--Identify strengths and weaknesses of "Mystery tweeter".

Note
The mystery tweeter is going to be used in a rs180 mtm (mym 180) project. I would like to learn more about the driver, since I know relatively little. In many ways, the driver appears similar to the dc28fs(silkie.)

Method
All drivers were tested in open air (no baffle)at approximately 4" using single channel sine sweeps. The loud tests were taken at an additional 10db of gain, relative to the medium setting. I am fairly confident that loud really is loud! The waterfall plots were taken at the "loud" setting.

I used the same "Medium" and "loud" settings for all 4 drivers(woofers/tweeters).

Challenges
I had a hard time deciding how to present distortion information since, the two different volume settings would throw the eye due to relative gain. In doubt, I decided to go ahead and use the same range for all tests. I am wondering how or if I should normalize the data in future testing.

I am also wondering if the loud setting was too loud! I have no way to know this. I used the same "Medium" and "loud" settings for all 4 drivers(woofers/tweeters).

Plots

DC28fs




Mystery




RS125




RS180

[Ludo]

Brian,

This is looking very promissing. Do you know what voltage levels the various loudness settings correspond to? Also, did you try moving the microphone further from the driver to see if the relative distortion levels remain unchanged and to make sure that you are not overloading the mic or the preamp?

Louis

[deLuxhifi]

Hmmmm. As a relative beginner at this I must say those 3d burst decay charts had never made any sense to me before. Whenever I viewed an individual burst chart for drivers I was looking at from manufactures spec sheets, it was like a Rorschat chart.

However, seeing a chart for a tweeter, a mid and another one for a woofer all together in one venue now I can see the differences and how each does its job at different Hz. A lightbulb went off...dim, but there's hope for me yet.

Well done.

Measurement gear almost seems a must.

How much would a reasonable testing rig cost, ballpark, depending on levels of quality?

I suppose there are other threads on this and I will do some searching.

Continue on the thread, I just thought I'd throw in my thanks for presenting a nice visual aid that really helped me a lot. Thanks.

[brianpowers27]

Do you know what voltage levels the various loudness settings correspond to? Also, did you try moving the microphone further from the driver to see if the relative distortion levels remain unchanged and to make sure that you are not overloading the mic or the preamp?

Louis

Ludo,

Those are both great suggestions. I misplaced my digital voltmeter and only have access to a cheap analog, at the moment. It does seem appropriate to standardize for voltage.

I am unfamiliar with the normal testing practices for this type of application. I have the D'Appolito - "Testing Loudspeakers" book setting on my desk. I just received it Monday. I will have to read more.

Delux - You need quite a few things to make it work reasonably well. I had most of these items prior to starting in the speaker hobby.
--Arta software
--Reference Amplifier
--Mic Stand
--Calibrated Mic
--Voltage Probe for reducing amp voltage back to the soundcard
--Nice Mic preamp
--Sound Card (Better if seperate from preamp) (M-audio delta 24/96)

[brianpowers27]

Tweeter Evaluation:
I am still trying to figure out what this mystery tweeter can do. This is the tweeter I won a month back on the guess that album cover. The fs is 1000hz and appears to be ferrofluid damped. The weight is close between the two with the extra density going to the mystery. THe mystery has a 4 hole metal faceplate.

It looks like the mystery beats the silkie in 3rd order at reasonably listening volumes but loses below 1000hz.

I am going to try crossing at 1800lr4.

[deLuxhifi]

Thanks Brian. I found Curt c's primer thread as well. Most helpful. A lot of stuff I was reading earlier is starting to sink in and make sense instead of generating a glazed film over my eyes like it did initially.

...and this is coming from someone who used to consider himself quite good at reading and interpreting graphs of various kinds.

[brianpowers27]

...and this is coming from someone who used to consider himself quite good at reading and interpreting graphs of various kinds.

Don't worry, I still look at these charts and feel somewhat unconfident when trying to draw conclusions.

I would like to note that there is much more that could be discovered via testing regarding the character of these speakers. I blieve that various configurations of IMD testing could yield good results. I am looking for input regarding IMD testing and would like to discuss this further.

[DDF]

Don't worry, I still look at these charts and feel somewhat unconfident when trying to draw conclusions.

I would like to note that there is much more that could be discovered via testing regarding the character of these speakers. I blieve that various configurations of IMD testing could yield good results. I am looking for input regarding IMD testing and would like to discuss this further.

You could consider the Phi spectral method developed by Jon. I like it as a one stop shop for a quick measure given the excitation is broad band and the spacing ensures a pretty clear picture of the distortion generated, but it does result in cluttered graphs than can be hard to interpret.

http://www.geocities.com/jonrisch/page10.htm

I took Jon's test signal and modified the spectral weights to be more representative of music, but its 6 of one half dozen the other. A flatter weighting gives an indication of peak spectral levels typically found in music whereas the weighted method is more an average over time.

Attached are graphs of signals I developed using NCH tone generator (excellent generator, tight tones and low spuriae). I could transfer these if you had an FTP site for me to drop to. I could also send some subjective test results from the AES using similar signal ("The Effect of Nonlinear Distortion on the Perceived Quality of Music and Speech Signals") send me a PM if you like and I can forward on.

I've also used the incoherency method advocated by MLSSA.

pg 213 @ http://www.mlssa.com/pdf/DRA-MLSSA-Manual-10WI-8.pdf

Mathematically its very elegant but the lack of a subjective anchor makes it hard to interpret (i.e. what is a "good" incoherency result?).

Dave

[Dennis H]

If those distortion plots are showing real SPL, you're WAY overdriving the Behringer mic at 140dB. The mic starts distorting when you get much above 100dB so you don't know if you're measuring driver distortion or mic distortion. And you should be using STEPS, not ARTA, for the best distortion measurements.

Edit: to eyeball mic distortion, leave the signal level to the driver the same and move the mic farther away. Driver distortion doesn't change so, if the measured distortion changes, it's the mic.

[tpremo55]

...And you should be using STEPS, not ARTA, for the best distortion measurements.

STEPS? Please elaborate. As you can imagine, a search for "STEPS" is pretty useless...

Thanks Dennis.

Sorry - I found it. "STEPS" is a component of ARTA Suite.

"STEPS - program for frequency response measurements with stepped-sine excitation."

[brianpowers27]

If those distortion plots are showing real SPL, you're WAY overdriving the Behringer mic at 140dB. The mic starts distorting when you get much above 100dB so you don't know if you're measuring driver distortion or mic distortion. And you should be using STEPS, not ARTA, for the best distortion measurements.

Edit: to eyeball mic distortion, leave the signal level to the driver the same and move the mic farther away. Driver distortion doesn't change so, if the measured distortion changes, it's the mic.

I'll have to try that. I am not sure what the real SPL for the mic is since I am using a SS preamp with a bunch of gain. I also considered the fact that my mixer may be adding some distortion. Ideally I would have a passive pre only between the amp and the soundcard.

I haven't tried Arta-Steps yet but I will have to do that soon.

[brianpowers27]

I would like to thank Dennis for pointing out the advantage of using Steps! The data is much more readable when the distortion is presented as a % of FR.

BTW: I test in 32 bit extensible with 96khz sampling.

Here are some new plots comparing the Silkie and the Mystery. Draw your own conclusions!

Mystery


Silkie

[Dennis H]

The data is much more readable when the distortion is presented as a % of FR

I like that better too. I also like STEPS showing the 5 and 6+ harmonics as they can be really obnoxious.