Speaker Testing

[Stub Mandrel]

I've got another thread about testing microphones.

 

I bought a cheap Behringer reference microphone and it seems to work pretty well using  it with REW to get a reference trace from my cheap Thomann powered speaker (EQ set flat, 4 10 second frequency sweeps at 0.5m in a large room with plenty of soft furnishings and curtains closed).

 

I asked a query there which has gone unanswered that I wonder of the likes of @stevie or @Phil Starr can answer.

 

REW also produces phase plots, and the one for the speaker is below. It shows several phase reversals, mostly well beyond hearing range.

 

One just above 100hz probably relates to the reflex port and is to be expected. 

 

Another is just below 2kHz and is accompanied by a dip in the SPL trace.

 

My question is: Does this make it likely my tweeter is wired back to front, and could reversing the wiring help eliminate the dip?

 

 

Any other observations welcome (yes I know this is not a very flat trace at all, although it sounds fine and works well as a monitor).

[Bill Fitzmaurice]

Phase shifts like that are normal, and don't matter, as you can't hear them. The dip and shift in the crossover region could be polarity related, the only way to know is to reverse the tweeter polarity and test again. It could also be time align related, which can only be fixed by moving the tweeter to get the acoustic centers of the tweeter and woofer the same. The shift at 100 Hz is probably port related, possibly indicating port tuning at 100 Hz. The 200 Hz dip could be from floor bounce. The best way to take measurements is outdoors, well away from boundaries, taking two measurements. The first is ground plane, with the mic a few cm above the ground. The second is with the cab on its back, the mic suspended above. This measurement will have a dip where the wave reflected off the ground meets the front wave at 180 degrees, the result of the baffle being 1/4 wavelength from the ground. The two are spliced to get the final result. You can do it in one step by digging a hole in the ground, putting the speaker in facing up, back filled with the baffle flush to the ground. It's how Roger Russell of McIntosh used to take measurements. 

 

When you take measurements 1/6 octave smoothing is preferred, as that's the limit of what you can hear.  

[Stub Mandrel]

Thanks  Bill lots of useful stuff there.

 

As far as I can tell the transducer in the horn is pretty well aligned with the main driver, so I'll experiment with changing its polarity to see what happens. 

Edited yesterday at 18:25 by Stub Mandrel

[Bill Fitzmaurice]

A bit of semi-interesting stuff is the design of the Altec A-7. The woofer horn had two purposes. One was increased sensitivity in the lower mids, but the other was to time align the acoustic centers of the drivers. The aligning procedure was to wire the HF driver reverse polarity, feed the speaker with a sine wave at the crossover frequency, then slide the HF horn on its sled atop the cab back and forth until the SPL reading was at a minimum. That placed the acoustic centers of the drivers correctly. Swap the polarity on the HF driver back and it was good to go. 

[Stub Mandrel]

Nice lateral thinking.

[Phil Starr]

14 hours ago, Stub Mandrel said:

Does this make it likely my tweeter is wired back to front, and could reversing the wiring help eliminate the dip?

That's a great question and what your measurements have uncovered is one of the problems that have cursed speaker design generally and crossover design in particular since they were invented. If not really a secret it is something buried inside any speaker with different speakers handling different parts of the frequency spectrum. It's approaching impossible to have the two speakers in phase at the crossover point and it is practically extremely difficult with a passive crossover to get the roll off of each speaker to be symmetrical. Inductors and capacitors induce their own phase shifts too so time alignment of two speakers at crossover is still really difficult. https://www.learnabout-electronics.org/ac_theory/ac_ccts_51.php Bill has pointed you to the Altec "Voice Of The Theatre" A7 which is a really interesting design. Large installed systems of the time frequently had horns mounted on a sled.

Active crossovers and DSP allow delay to be brought into the crossovers hence the rash of 'Firphase' into PA speakers, this is the problem it is addressing.

 

Practically then you can pretty much always see a frequency anomoly at crossover. the designer has the choice of a blip or a dip and crossover distortion is audible in critical listening tests. Usually it shows up quite clearly and affects female voices in particular. We are really sensitive to any change in voices and the frequency of crssover is usually well within the range of female voices. 

 

Practically a dip at crossover is usually less obtrusive than a blip. Try reversing the wiring and you should see a change in your measurements and hear differences in voices and instruments in that area. Acoustic guitar often changes too but I use classical music for testing this sort of thing. Even tiny changes in a crossover can make an instrument pop out in a recording. You can choose your distortion  mostly I prefer the dip.

 

@stevie is the expert in this and the detail he goes into with his crossovers is extraordinary. It's really worth getting hold of a copy of the Loudspeaker Design Cookbook by Vance Dickason if you want to read up on all this, and much more.

[stevie]

The phase at 2kHz (and elsewhere) isn't a phase reversal: it's how the phase is represented when it passes 180 degrees. To see what's actually happening with the phase, you need to compensate for time of flight. Right-click on the phase curve and select Estimate IR Delay (on the latest version of REW) - or press Ctrl+Alt+E. Then have another look at the phase graph. Also, the dip at 2kHz is not typical of a phase reversal dip - unless you're using smoothing, of course, which would tend to fill in the dip.

 

 

[Bill Fitzmaurice]

5 hours ago, Phil Starr said:

 It's approaching impossible to have the two speakers in phase at the crossover point and it is practically 

 'In phase' implies that one needs the phase response of the drivers to be the same. That is pretty much impossible. But it doesn't matter. All that matters is that they're not 180 degrees or close to that apart at or near the crossover frequency, and as such a simple polarity reversal of one driver is the usual fix. As for crossover design, that used to be a seat of the pants job. You'd start with a calculated component value, but to get it perfect required repeated component swapping and testing to dial in the final result. That hasn't been the case for quite some time, thanks to software like XSim, which unlike traditional calculations doesn't rely on the nominal impedance of the drivers for the calculation but instead uses the full frequency response (FRD) and impedance sweep (ZMA) plots to get an exact result. The only issue with that is finding FRD and ZMA files for the drivers, which are only gradually finding their way onto manufacturer driver data sheets, though you can measure those yourself with the necessary gear. Crossover modeling software is to crossovers as speaker modeling software is to speakers. 

[Phil Starr]

We will have to agree to differ on this Bill. You can see the frequency irregularities around the crossover point in pretty much any frequency plot I've ever seen or measured and the irregularities are audible as well as measurable.  Without wanting to get into a philosophical debate about the nature of reality IMO if something is measurable and audible it is real.

 

There's no doubt the software is getting better and will get better still but again IMO it isn't there yet. The measured plots still don't exactly match the computer's modelled responses and most of the big designers are going to use the software to get in the ball park and still use iterative measurements to check and hone their designs. Again I suppose it depends upon what you mean by "an exact result" it may be good enough but I'm a scientist not an engineer so exact is a very big statement.

[Bill Fitzmaurice]

With respect to using XSim, it's not a modeled response per se, as it uses the measured response and measured impedance of the drivers to arrive at the result. All it does calculation wise is to plug in the effect of the crossover components, so it's as accurate as the tolerance of the components used. Having used it myself I can attest to it's accuracy, which is also better than a traditional calculator as it can account for the ESR of both the inductors and capacitors. 

[Stub Mandrel]

This is all fascinating.