Measuring Loudspeaker Impedance vs Frequency

[jez_145]

I'd be interested to hear any suggestions on how to measure the impedance versus frequency characteristics for a typical loudspeaker drive unit i.e. the actual test gear you would use to obtain the data so as to be able to plot impedance versus frequency across the full audio range 20Hz to 20kHz. I'm talking about using gear we might have at home - not multi-thousand pound instruments from the likes of Bruel & Kjaer or Audio Precision!

I'm also interested to know how the impedance vs frequency characteristics for a given speaker drive unit/enclosure may be affected when driving it at different power levels. For instance, when using a typical signal generator output directly as a variable frequency source as compared to a more heavily amplified frequency source (causing greater excursion of the speaker cone).

As we know, measuring the DC resistance of the voice coil can often give a reasonable idea of a speaker's nominal impedance. For instance, a DC resistance of 3.5 ohms may indicate a 4 ohm impedance, 7 ohms DC resistance may indicate 8 ohms impedance and so on. Of course, in practice, the actual impedance of a speaker can vary enormously throughout the audio frequency range and may resemble something like that shown in the attached plot. In this example, a low frequency peak occurs at around the 180Hz mark, due to the mechanical resonance of the driver unit. The size of the peak and the frequency at which it occurs is influenced not only by the mechanical characteristics of the driver (its mass, suspension arrangements etc.) but also on the design of the enclosure containing the driver (if any). After levelling off, the impedance then starts to rise rapidly after, in this case, the 4kHz point, this time due to the reactance of the voice coil which is itself inductive. This is what I want to be able to plot.

In his book, "Building Valve Amplifiers" (Newnes/Elsevier 2007) Morgan Jones suggests setting up a potential divider using an "oscillator" in series with a fixed resistance and the loudspeaker and then measuring the voltage attenuation in the setup shown in the second picture. Here, it is assumed that the oscillator's output resistance is constant with frequency. Then the potential divider equation is used in reverse such that:-


Z loudspeaker = R oscillator/((V loudspeaker / V oscillator (open circuit)) - 1)

where V oscillator (open circuit) is the output voltage of the oscillator without any loading.


As an alternative to using voltage measurements and deriving the impedance at various frequencies using the above equation, how about using an LCR bridge ? I have a Marconi TF 2700 universal bridge for instance. I admit that I've only ever used the internal 1kHz oscillator on this unit for occasional resistance and capacitance measurements. However, I'd be interested to know how to use the external AC input on this instrument, in conjunction with a signal generator (and possibly via an isolating transformer), to take measurements at various frequencies in the 20-20kHz range.

Any ideas/discussion welcomed,

Many thanks.

Jerry

[trobbins]

I have used REW and soundcard. There is a tech note on speaker impedance measurement using a simple jig.

[Herald1360]

If you simply take the ratio of the voltage across the speaker to that across the 600R and multiply that by 600R, it doesn't matter what the actual sig gen voltage or output impedance is as long as the on load sig gen voltage is high enough to give a readable level across the speaker when its impedance is low. (At low/mid frequencies the speaker voltage may only be 1% or so of the 600R voltage).

This only gives the magnitude of the impedance, not the full picture.

A 'scope may be better than a meter with unknown frequency response to take the measurements- it'll only be accurate to maybe 5% but at least that's a reasonably well known accuracy.

[Diabolical Artificer]

I've done it using a sig gen, frequency counter and DMM using your circuit, but measuring the voltage across the resistor to find current, keeping input voltage constant. Think that's how I did it, bit early and not had coffee yet.

Or use REW as Tobbin's said.

Andy.

[Craig Sawyers]

If I were doing this, I'd use ARTA. I have this for measuring active filters, although it does much, much more. There is a module called LIMP (Loudspeaker IMPedance) which seems to do what it says on the tin.

http://www.artalabs.hr/index.htm

Full license is 79Euro, but you can run in "demo mode" for free in which you can't save or load files.

Craig

[GrimJosef]

When you say you want to measure "the impedance versus frequency characteristics for a typical loudspeaker drive unit" do you mean for the drive unit in isolation, or are you thinking of doing this with the driver mounted to its cabinet/baffle ? In either case you may have to worry about the mechanics of the measurement as well as the electronics. If, for example, you were to sit the driver on its back on a flat bench the back-reflected wave from the bench might affect the cone motion to the point where the speaker's impedance was altered.

I'd also reinforce Chris's point that dividing the amplitude of the voltage by the amplitude of the current will only give you the amplitude of the impedance. If you're interested in the way the speaker interacts, say, with a crossover or an amplifier then it's very handy to know the relative phase of the voltage and current signals too. Some sort of 2-channel oscilloscope (either a stand-alone unit or a computer-based one) is very handy in this respect.

As far as the dependence of the speaker's behaviour on power level goes, that is essentially a question of linearity - if the behaviour is truly 'linear' then quantities like impedance won't vary with power. If the behaviour is not linear then an inevitable consequence is signal distortion. I believe that these days distortion in good quality drivers can be down at the percent level although it does rise spectacularly when you get close to the driver's power limit. It is worth saying, of course, that a) driving speakers with very high levels of cw sine waves can damage them and, before that, may lead to their behaviour changing as the voice coil heats up and b) you may well not be able to stay in the same room as a driver producing a great deal of cw audio power. At the other extreme i.e. very low signal level all should be well as long as there's no fault leading to 'stiction' in the cone motion.

Cheers,

GJ

[Craig Sawyers]

All of that is absolutely bang on. Moving coil loudspeakers are complex in their behaviour.

For anyone wanting some light bedtime reading, Wolfgang Klippel is the world authority on this, and his software is used by I believe by just about every driver manufacturer.

https://www.klippel.de/know-how/literature/papers.html

Craig

[trobbins]

REW has an active forum for all things speakers, and has a detailed TS parameter worksheet to link to the impedance plots, and latest FDD modeling. And its free, and easily connects to java and ASIO drivers.

[PJL]

The 600 ohm measurement method is fine. It is important the sine wave source and amplifier have very low distortion levels.

[jez_145]

Well chaps, you certainly didn't let me down and thank you very much for your replies. I was about to do a few measurements/experiments with my function generator, DMM and scope (still will do) but thought I'd throw the question open just to see what other methods came up. The REW and ARTA programs are a revelation to me and I've now downloaded them to experiment with. Coincidentally, I recently purchased a very good quality USB audio interface for my PC and I can see I'll be putting this to even more good use very soon.

Thanks also Craig for the link to the writings of Prof. Dr. Wolfgang Klippel. Some very interesting reading there.

Jerry