JBL M552 CCBFREQ measured!

[frank23]

Due to some (clever) thinking and linking up my macbook line in with my stereo, I have made some analysis with Audacity that I wanted to share with you.

What I did was to measure a full range white noise frequency spectrum from:
1) line out on my cd player
2) high output on my M552 which is wired for CCBFREQ and was crossed over at 200Hz

By comparing these using the spectrum analysis function of Audacity you can see the CCBFREQ effect from the M552. I am not too sure about the dB scale. That is why I crossed it at 200Hz to be able to see if I can find the 24dB/oct. This looks correct (see the about 24dB difference between 100Hz and 200Hz), so does this mean the CCBFREQ is really only about 6dB between 2000Hz and 17500Hz?

first the full range line signal from the cd-player output:
http://i200.photobucket.com/albums/aa253/frankvcp/whitenoiselineoutcdplayer.jpg

then the fullrange signal (hmm, actually it looks like it was crossed at a low frequency, but just look from 200Hz upwards) from the high output of the M552 with CCBFREQ:
http://i200.photobucket.com/albums/aa253/frankvcp/whitenoisethroughm552cccorrection.jpg

then the signal crossed over at 200Hz from the high output of the M552 with CCBFREQ:
http://i200.photobucket.com/albums/aa253/frankvcp/whitenoisethroughm552cccorrectioncr.jpg

I'd love to hear what you think of it!

[1audiohack]

I think your very close. I measured the T.F. of one of my M552's set on CC comp for the 2360 so I could duplicate the curve in a DriveRack. The EQ spread I wrote into the DBX was -4.5 to + 3dB, as seen on the DBX screen at the bottom, 7.5dB total. The compared curve is shown below, the white line is the M552, the yellow one is from the DBX.

I had the M552 crossed about 1kHz and I neglected to switch the TEF screen from dB Pascal to dBV so that scale is a guess.

Good to see someone in else in the discovery stage! Keep it up and have fun.

All the best,
Barry.




http://audioheritage.org/vbulletin/attachment.php?attachmentid=37103&stc=1&d=1234000900



http://audioheritage.org/vbulletin/attachment.php?attachmentid=37102&stc=1&d=1234000200

[Zilch]

http://www.audioheritage.org/vbulletin/showthread.php?p=135168&#post135168

[Robh3606]

The cut and boost used for the 2360/2380 horns is published in the owners manuals for the M552/M553. Just download the manual. Unity gain is 5k for both curves.

http://www.jblpro.com/pub/obsolete/M552%20Owners%20Manual.pdf

Rob:)

[frank23]

The cut and boost used for the 2360/2380 horns is published in the owners manuals for the M552/M553. Just download the manual. Unity gain is 5k for both curves.

http://www.jblpro.com/pub/obsolete/M552%20Owners%20Manual.pdf

Rob:)

Hi Rob,

I have the manual, but that does not give you any info on the shape of the curve. Looking at zilch's link above, the FF compensation cards for other JBl crossovers have a bump lower down. In the M552 manual the only difference in specs (page 19) between the CC and FF is that CC is -5dB at 2000Hz and FF is -4dB.

I wanted to see the curves. I will measure CC and FF side by side coming weekend.

[boputnam]

Novel approach. Interesting.

However...
...(see the about 24dB difference between 100Hz and 200Hz...that looks like -18dB/octave to me.

[pos]

The problem with audacity is that you need white noise instead of pink noise.
Or you have to export the data and apply some compensation.
I am reluctant puting high level white noise into my full range speakers...

[boputnam]

The problem with audacity is that you need white noise instead of pink noise.I was wondering the same thing. I've never measured White Noise. Pink Noise (filtered White Noise) is a more relevant reference - it's use might change these observations.

[pos]

from the first picture it looks like frank23 is using white noise, wich is what render a flat curve with audacity.
Using pink noise gives you the classic descending curve, and the soft provides no mean of compensating it to flat.

[boputnam]

from the first picture it looks like frank23 is using white noise, wich is what render a flat curve with audacity.Yes, he stated he is using it.


What I did was to measure a full range white noise frequency spectrum from

I do not know how it would affect his plots of the CCBFREQ and the 552 crossover.

[frank23]

I am reluctant puting high level white noise into my full range speakers...

I measured the output of the M552, so not the speaker output.

[frank23]

Yes, he stated he is using it.



I do not know how it would affect his plots of the CCBFREQ and the 552 crossover.


I think this method is ok with Audacity. I wil measure some other test signals next time that I know the dB difference of for sure to be able to calibrate the Audacity scale. Maybe it is correct as I've measured it, maybe not, I'll know next time.

[pos]

It is correct. Here is what a pinknoise looks like in audacity.

[pos]

Makes me think:
You could also use a sweap noise (limited to 23s...). But as for the white vs pink, you would have to use a linear sweap!
I whish Audacity had a correction option to flatten the "pink" log spectral balance...

[grumpy]

not main topic, but...

I haven't played with Audacity's analysis function much, but it appears
there is a "chirp" generator that -could- be used to good effect for such
measurements. (tried it briefly using v1.3.8beta on a mac, onboard
speakers and mic). Might have to work with it a bit, and perhaps get
some expert feedback from an Audacity-centric forum, but a here's a
brief comparison using Audacity (100-20000Hz log sine sweep) vs.
Fuzzmeasure doing essentially the same thing (measuring mac laptop
speakers and mic, -not- M552/3):

[Zilch]

It is correct. Here is what a pinknoise looks like in audacity.http://en.wikipedia.org/wiki/Colors_of_noise

http://en.wikipedia.org/wiki/Pink_noise

http://en.wikipedia.org/wiki/White_noise

[pos]

Most measurement software, like smaart for example, calibrate their curve to show a flat curve for a pink noise spectral distribution.

This is better because it is the "normal" distribution in music. If you inject high level white noise into your speaker you may overload the tweeter (same symptome as the "not enough power" thingy), so white noise is really not the best for speaker measurement.

That was my point in post #7

[ratitifb]

Noise distribution vs frequency is only depending on the way you are performing the analysis ;)

For example, pink noise distribution is flat with 1/n-ieme octave analysis (constant percentage bandwidth (CPB) resolution ie delta(F)/F = constant) while the distribution of the same pink noise is decreasing by 3dB/octave with FFT analysis (constant bandwidth resolution ie delta(F) = constant)

[frank23]

I have done some more measurements on my m552 and it turns out the vertical scale of the Audacity analyser is correct.

I measured a 1001Hz tone at 0, -20, -40, -60 dB (from the denon test cd) and this was exactly replicated in the Audacity analyser graphs.

So, I'd say the above graphs I made are pretty accurate.

So I did some more analysis. I built myself a 3-way (see in the DIY section) for which I use the m552 in mono 3-way mode. The following is the output on the "mid" channel when crossed over at 200Hz and 4000Hz, according to the scales on the front of the m552. In reality it looks like 200Hz and 3000Hz ! Also 2000Hz on the scales turned out to be 1500Hz in reality.

http://i200.photobucket.com/albums/aa253/frankvcp/m5523waymonomidout200Hz4000Hzonscal.jpg

So before you say that "it doesn't sound good at xyz", you should probably try and measure the output as in this case it differs substantially.

[frank23]

Most measurement software, like smaart for example, calibrate their curve to show a flat curve for a pink noise spectral distribution.

This is better because it is the "normal" distribution in music. If you inject high level white noise into your speaker you may overload the tweeter (same symptome as the "not enough power" thingy), so white noise is really not the best for speaker measurement.

That was my point in post #7

ok, I get it, seems sensible to not load the tweeter with the same "full" signal as the woofer

[pos]

you can also use a linear sweep, generated with this:
http://www.satsignal.eu/software/audio.html#SweepGen

(you have to choose the "linear" option, as well as limit the duration to something like 20s as audacity is limited to 23.8)

[frank23]

I have bought an M553 for my active 3-way project, here are some audacity graphs comparing the FFBREQ and CCBREQ settings.

The first graph is the mid output with the lower crossover at 18Hz and the upper at 450Hz. It is just for level comparison:
http://i200.photobucket.com/albums/aa253/frankvcp/m553_mid_0dB_18Hz_450Hz.jpg

The second graph is the high output with the crossover at 450Hz and CCBREQ:
http://i200.photobucket.com/albums/aa253/frankvcp/m553_high_0dB_450Hz_CCBREQ.jpg

The third graph is the high output with the crossover at 450Hz and FFBREQ:
http://i200.photobucket.com/albums/aa253/frankvcp/m553_high_0dB_450Hz_FFBREQ.jpg

[frank23]

Pfff, three years on, I am still doing it... Here some more measurements I did today after finding out how to compensate Audacity for pink noise. You take a full spectrum and take that as a reference in Excel, instead of just publishing the Audacity curves. Then in Excel you get this for the M552 and M553, both on the 2380 CD compensation settings:

https://lh4.googleusercontent.com/-O45N8HjWcrI/UF8UrgLGnjI/AAAAAAAAOKQ/Biulmnb6ZCA/s800/m552%2520m553%25202380_setting.jpg

I adjusted all the measurements to 0dB where they should be by adding or substracting xdB. You can see that the crossoverpoint around 175Hz is not perfect. The M552 seems better. Furthermore there is a strange wiggle in the M553 in both of the mid curves at 100Hz (blue and purple lines). But this is an old copy of the M553, I'll have to measure the one I am actually using.

[boputnam]

As it was three years ago: Interesting...

[Oldmics]

Are you guys inducing a load on the output bands for these measurements?

Oldmics