The primary difference between an SM57 and the SM58 is the windscreen, the 58 having more "pop" and wind noise immunity. The SM57 cartridge is slightly closer to the screen than the SM58, so the proximity effect and output level is slightly greater when "kissing" the source.
The proximity response of a mic of this type would make distortion levels of another source difficult to determine, as the fundamental tone may raised +15 dB (or more) compared to the harmonics, depending on distance, around a 20% distortion difference.
Condenser and back electret mics generally have a hard SPL limit where the output actually clips, distortion may go from a fraction of a percent to over 100% in a just a few dB.
Dynamic mics generally won't reach a hard limit, but their transformer saturation and diaphragm asymmetry does progressively add harmonic distortion at higher SPL levels, all a part of the "sound character" that still makes them a staple for recording.
As far as the signal to noise ratio between mics, if the distortion components of the driver under test are so low as to be masked by the noise of a omni condenser mic with a decent S/N, they probably won't be noticed in a listening test.
Art
If you use Matlab you could try this: https://www.diyaudio.com/community/threads/loudspeaker-distortions-measurement-in-matlab.388032/
Interesting!
I was 'forced' to use Matlab in school 25yrs ago, but I did not like it, and whatever I learned has been forgotten.. Had I seen a use in audio, I might actually have learned a lot more. I guess Matlab has evolved at the same rate as audio SW during this time
I remember buying some measurement SW I read about in speaker builder back then, and getting it on a floppy disk in the mailbox SW was pretty useless, using a very short pulse as output, so there was no bass energy in the signal, and the measurement became extremely sensitive to external noise. Not everything was better in the past
Thinking about it, I could probably get matlab installed on my work laptop..
I was 'forced' to use Matlab in school 25yrs ago, but I did not like it, and whatever I learned has been forgotten.. Had I seen a use in audio, I might actually have learned a lot more. I guess Matlab has evolved at the same rate as audio SW during this time
I remember buying some measurement SW I read about in speaker builder back then, and getting it on a floppy disk in the mailbox SW was pretty useless, using a very short pulse as output, so there was no bass energy in the signal, and the measurement became extremely sensitive to external noise. Not everything was better in the past Thinking about it, I could probably get matlab installed on my work laptop..
My guess is that this explains 95% of the failure with education - creating that specific interest to the student (or rather - utterly failing to do so).
I'm not an engineer or mathematician, but my first thought of IMD measurements is to measure and compare multidriver loudspeakers. Single driver measurements in nearfield perhaps can't give any more info than a normal HD measurement.
I have both ARTA and REW (Thank you JohnPM!), but I use only REW because it has friendy UI and I have just a simpe USB mic (Umik-1). Main problem is to have guiet and reflection-free environment.
There is lots of IMO-grade discussion about IMD and Doppler of various speaker constructions and xo types. Still I haven't noticed any kind of general trend between types. And my guess is that HD shows obvious and audible problems well enough.
I have both ARTA and REW (Thank you JohnPM!), but I use only REW because it has friendy UI and I have just a simpe USB mic (Umik-1). Main problem is to have guiet and reflection-free environment.
There is lots of IMO-grade discussion about IMD and Doppler of various speaker constructions and xo types. Still I haven't noticed any kind of general trend between types. And my guess is that HD shows obvious and audible problems well enough.
Last edited:
GNU Octave is roughly equivalent to matlab and free software https://octave.org/Interesting!
I was 'forced' to use Matlab in school 25yrs ago, but I did not like it, and whatever I learned has been forgotten.. Had I seen a use in audio, I might actually have learned a lot more. I guess Matlab has evolved at the same rate as audio SW during this time
Thinking about it, I could probably get matlab installed on my work laptop..
https://en.wikibooks.org/wiki/MATLAB_Programming/Differences_between_Octave_and_MATLAB
FYI Member @mbrennwa has published lots of audio related scripts that work with Matlab / Octave, including calculation of Gedlee distortion metric
https://github.com/mbrennwa/mataa/blob/master/mataa_tools/mataa_measure_GedLee.m . There is manual in the same Github repository with examples how to use, just browse to find it.
Last edited:
I have a feeling that the mic distortion and background noise are the main limiting factors here, looking at the measurements I posted. Looking at JC's video posted here on the previous page where he compares with the Shure mic, is a clear indication of mic distortion being a problem. Even the B&K I'm using is probably masking with it's own distortion if the driver is performing well. This is a learning experiment for me, and I have already learned that driver distortion is lower than I thought, or should I say I was led to believe by many of the measurements I have seen. One example would be Troels distortion measurements, since I measure significantly lower distortion than he did on the same driver, even if I'm using multitone, so IMD is added to the mix.
I guess there can be reasons to measure both single drivers and complete speakers.
Measuring the single driver could give some guidance to how to use it, and measuring the complete speaker to see how well the end result is (and what you are hearing when you listen to your speakers). Maybe tweaking DSP filters and comparing would be the best when measuring single drivers?
I guess there can be reasons to measure both single drivers and complete speakers.
Measuring the single driver could give some guidance to how to use it, and measuring the complete speaker to see how well the end result is (and what you are hearing when you listen to your speakers). Maybe tweaking DSP filters and comparing would be the best when measuring single drivers?
Check out lrisbo writings on the forum and tech notes and blog at Purifi website. From top of my head he has said and demonstrated doppler distortion associated with cone excursion is less audible compared what amplitude modulation can happen with the same excursion. Position dependent force factor BL(x) makes amplitude modulation, also Sd can vary with excursion making amplitude modulation. They've written somewhere in the blog/papers its often hard to discern sources of measured distortion and then write what they are, basics how to drop loudspeaker driver THD number by a lot, interesting stuff. It can be argued though, what is audible and what is not. I'm sure small two way speaker benefits very good woofer, but not sure if big three way speaker with 15" bass will as much, as distortions ought to be much less anyway, perhaps below audibility. Certainly big speakers are much fun listen to
Perhaps very good woofers make small speakers fun too, never heard though.edit. here is comparison of phase / amplitude modulation https://purifi-audio.com/2019/12/07/amfm/ , perhaps lrisbo never said amplitude modulation would be more audible and its just my imagination / memory
Anyway, the site is good resource to familiarize with if loudspeaker distortion is of interest.
Last edited:
Speaking of signal generators for arbitrary signals, free software, Matlab (or GNU Octave), and distortion analysis: MATAA does it all (and more).
https://github.com/mbrennwa/mataa
https://github.com/mbrennwa/mataa
REW is also good with signals, including multitones!
German reviewers use often Klippel systems, and often show multitone measurements too. But I can't get a grip of those, various levels...
https://www.fidelity-online.de/test/lautsprecher/lautsprecher-messungen/
https://www.soundandrecording.de/thema/studiomonitore/
Ebenso erwähnenswert wie die Maximalwerte ist der gleichmäßige Kurvenverlauf. Schwachstellen, wo die Kurve einbricht, gibt es beim KH 420 nicht. Die zweite Messreihe wurde zum Thema Intermodulationsverzerrungen mit einem Mittlungspegel von 85 dBA in einer typischen Hörentfernung von 4 m unter Freifeldbedingungen durchgeführt. Der Spitzenpegel bei dieser Messung, ebenfalls in 4 m Entfernung, betrug 100 dB, entsprechend 112 dB in 1 m Entfernung. Als Testsignal wurde ein Multisinus mit 60 Anregungsfrequenzen und einer Gewichtung nach EIA-426B für ein mittleres Musiksignal genutzt. Das Signal hat einen Crestfaktor von 12 dB. Die Grafik aus Abbildung 15 zeigt dazu das Spektrum des Anregungssignals (grüne Kurve), das gemessene Spektrum des vom Lautsprecher abgestrahlten Signals (rot) und die daraus extrahierten Verzerrungsanteile (blau). Beides wird sowohl mit einzelnen Spektrallinien wie auch in 1/6-Oktavbandbreite aufsummiert dargestellt. Der Gesamtverzerrungsanteil (Harmonische und Intermodulationen THD + IMD) liegt A-bewertet für den Abhörpegel von 85 dBA Leq in 4 m Entfernung bei sehr schön niedrigen -34 dB, entsprechend 2%. Auch hier ist wieder gut zu erkennen, wie die ohnehin schon geringen Verzerrungen oberhalb von 500 Hz noch mal deutlich geringer werden. Ein Dreiwege-System kann hier seine grundsätzlichen Vorzüge ausspielen, wo die Mitten nicht mehr durch die großen Auslenkungen im Bass beeinträchtigt werden. Hinzu kommen die Verzerrungsarmut und die hohe Sensitivity der beiden Kalotten.
German reviewers use often Klippel systems, and often show multitone measurements too. But I can't get a grip of those, various levels...
https://www.fidelity-online.de/test/lautsprecher/lautsprecher-messungen/
https://www.soundandrecording.de/thema/studiomonitore/
Ebenso erwähnenswert wie die Maximalwerte ist der gleichmäßige Kurvenverlauf. Schwachstellen, wo die Kurve einbricht, gibt es beim KH 420 nicht. Die zweite Messreihe wurde zum Thema Intermodulationsverzerrungen mit einem Mittlungspegel von 85 dBA in einer typischen Hörentfernung von 4 m unter Freifeldbedingungen durchgeführt. Der Spitzenpegel bei dieser Messung, ebenfalls in 4 m Entfernung, betrug 100 dB, entsprechend 112 dB in 1 m Entfernung. Als Testsignal wurde ein Multisinus mit 60 Anregungsfrequenzen und einer Gewichtung nach EIA-426B für ein mittleres Musiksignal genutzt. Das Signal hat einen Crestfaktor von 12 dB. Die Grafik aus Abbildung 15 zeigt dazu das Spektrum des Anregungssignals (grüne Kurve), das gemessene Spektrum des vom Lautsprecher abgestrahlten Signals (rot) und die daraus extrahierten Verzerrungsanteile (blau). Beides wird sowohl mit einzelnen Spektrallinien wie auch in 1/6-Oktavbandbreite aufsummiert dargestellt. Der Gesamtverzerrungsanteil (Harmonische und Intermodulationen THD + IMD) liegt A-bewertet für den Abhörpegel von 85 dBA Leq in 4 m Entfernung bei sehr schön niedrigen -34 dB, entsprechend 2%. Auch hier ist wieder gut zu erkennen, wie die ohnehin schon geringen Verzerrungen oberhalb von 500 Hz noch mal deutlich geringer werden. Ein Dreiwege-System kann hier seine grundsätzlichen Vorzüge ausspielen, wo die Mitten nicht mehr durch die großen Auslenkungen im Bass beeinträchtigt werden. Hinzu kommen die Verzerrungsarmut und die hohe Sensitivity der beiden Kalotten.
Last edited:
As always if I am asked for mic recommendations, I say IsemCon-7150 and Earthworks. They are worth every penny, and it's great to work with tools that work well.
@Juhanzi As i understand it from the text, green is the FR of the signal, red is the FR output of the speaker, and blue is distortion. But I don't really see corresponding SPL in the graph as they mention in the text. I also don't understand that there are blue bars and blue 'grass' for distortion..
OK, now I see distortions. Another problme is why they use so different spl levels... Is it to set max. distortion to 2%, then to see the "clean" output capacity (in spl)? As benefit of this 3-way speaker they mention the low midrange IMD.
In near future I don't have time to set up my own measurement system. My "lab" is too small for speaker measurements, other than nearfield.
In near future I don't have time to set up my own measurement system. My "lab" is too small for speaker measurements, other than nearfield.
It looks like Soundeasy has a function that can do the sort of measurement we've been talking about. I'd never used it before. But it has 3 generators, where one can be set to single frequency to set excursion, and the second can sweep. The 2nd,3rd,5th order IMD components can be tracked and displayed, I assume, as a single line like THD%. You can also set the third generator to track the sweep generator at a fixed frequency or multiple. So a sweeping multitone with a fixed low frequency to set excursion. I need to look into this soon and see if meets my needs.