measuring loudspeaker power on pink noise

[danmcb]

Say I drive a speaker with pink noise, and I can accurately measure the voltage across it, and the current through it, with a sense resistor. (I have a test set that will measure true RMS of noise with full bandwidth, and it has floating differential inputs, with 100k Zin).

Now the electrical  power consumed by the speaker is just the product of V and I - less a correction factor of about 3dB because the signals are not coherent, right?


(EDIT : I think I'm wrong  about the 3dB. That's for *summing* non-coherent signals. It's just a simple VxI calculation, assuming that "true RMS" is for real.)

[b_force]

Just have a read on crest factors.

[danmcb]

I did, and I know what crest factor is. But how does it affect power measurements? thanks.

[BrianHG]

Now the electrical  power consumed by the speaker is just the product of V and I - less a correction factor of about 3dB because the signals are not coherent, right?

Calculating power consumption as a whole at the speaker input, measuring the voltages on a measuring device capable of working with reading the true RMS levels of the noise may work, but, how do you come to the -3db & the signals not being coherent?

Your are trying to measure the actual power consumed by the speaker, not correct for it's internal load & efficiency spectrum, correct?

[b_force]

I did, and I know what crest factor is. But how does it affect power measurements? thanks.

It's a little difficult for loudspeakers.
Because the DC resistance (Re), changes when the speaker gets warmer.
Which also changes other parameters like the Qes (electrical Q) and Vas.
This is also known as 'power compression'.
In the end it changes the total response of your loudspeaker.

It all depends what you really want with it.
It's not completely clear to me if your question is just purely theoretical or if you actually want to measure the power?
And if you actually want to measure it, is it just for research, (power) limiting or even on the fly correcting of certain errors?
They all have a different approach.

[snarkysparky]

I don't see any way other that forming the instantaneous product of voltage and current and averaging that signal.

The trouble would be the speaker impedance not being constant across frequency.

pipe voltage and current into an analog multiplier IC.   Determine the correct scalings and output units.

[BrianHG]

The trouble would be the speaker impedance not being constant across frequency.

If you have the money and will, I can sell you a pair of Clements RT-7 speakers with a flat 8 ohm response throughout 0hz to 100Khz guaranteed with any signal as it's crossover is just a precision tunes R-C filter & the woofer's load response is balanced due to the intelligent cabinet design.  The tweeter is a transformerless planar ribbon which also exhibits a flat resistive load across it's frequency range.

What I can't guarantee is how much the impedance slightly increases as the components heat up.

Again, to the OP, do you want to just measure the raw power consumed by your speaker for evaluating power consumed by said speaker, or, are you trying to evaluate the speaker's response to the power being fed in, this will alter the approach to your test.

With the method you placed in you OP, you should be able to measure overall power consumed by your speaker is your source measuring HW is sampling everything in parallel simultaneously.  This means 2 differential channels sampled together.  In my circumstance, I might use a 8 ohm resistor in series as my current shunt with a 4x voltage output amp & do the math, however, this allows some woofer and crossover circuitry to interfere with the linearity of the source pink noise.

What king of accuracy will you need?
Something like within 1 watt over a 100 watt range?

[Audioguru]

The frequencies and levels of pink noise are random so the average power used by the speaker will be accurate if the amplifier is not clipping the peaks.

[MagicSmoker]

I know there is an industry trade group standard for measuring speaker power with pink noise (AES, maybe?) but IIRC it assumes a constant impedance over frequency and a certain peak to average ratio in the noise (ie - crest factor). So, most likely a case of Garbage In, Garbage Out.

An entirely different approach that will likely be more accurate is to measure the input power to the amplifier, as long as it isn't operating in Class A... For Class AB types the output power will be over-estimated due to the standing bias current and various losses, but these errors can be zeroed out by comparing with a known, purely resistive load. A Class D amplifier, however, will demand input power that very nearly tracks the real component of output power so should be ideal for measuring power into a complex reactive load like a speaker (NB - the reactive component of output power just sloshes back and forth between the speaker and the DC power supply capacitance, hence you would need to measure V, I and the phase angle displacement between the two to get a true reading a power if measured at the speaker).

edit: grammar

[danmcb]

3dB was an error on my part. If you SUM two noise signals, they sum at ~3dB not 6. Not relevant to power calculations.

[danmcb]

The reason for using pink noise is, as someone pointed out, to get a spread of all frequencies - thus averaging away variations in impedance and also the worst of the effects of the room.

I am looking to make real world comparative measurements of two speakers used in a high end 100V paging system. They have an existing type which has certain speaker tap levels (1W, 2W, etc) and a new type with different tappings. I wanted to compare real world measurements to help them provision speaker  placements (that means finding out how an "8W" tap on speaker A compares in volume to a "16W" tap  on speaker B) and also just get some feel for how they load the amplifier.

So far the readings and calculations that I made seem to give sensible results, and things are as expected. There are margins of error in the system design, so we don't expect or need 1% accuracy - just checking that there are no nasty anomalies waiting to be discovered when it is too late to change anything.

[b_force]

If you just gonna make a relative comparison, it really doesn't make a difference.

If you would like to compare with competitors....
Just forget it. I have worked for many years in this field and specs are just sucked out a very big thumb.
Ideal situations, ideal specs on ideal loads etc (especially for the amplifiers!!)
130dB, with 1000W, out of a 10 inch speaker, riiiight..... 

The reasons is that there are no decent standards (except for the AES noise one, but that one is very limited and old).

Even that doesn't say the whole story, because you also need specs like directivity and other things to get the full picture.
On the other hand, companies also full around with electronics specs (the input specs for ADCs for example)

[Benta]

Perhaps I'm missing something here, but...

Measuring the current and voltage going into the speaker (whether it's pink noise or a frequency sweep or something else) will tell you how much electrical power is sent into the speaker.
But it tells you nothing about how efficiently the speaker converts this to acoustic power vs frequency.
So you'd have to optimize the electrical power transfer to be as linear as possible. Then you'd need to set up an acoustic test to measure SPL. And then modify the electrical power transfer to reflect the inverse of that function.

Or?

[danmcb]

b_force : it is not about comparing with competitors - I know better than to indulge in specmanship in the audio industry. It is for two purposes, in an actual existing system in a large building.

1) they are gonna hang these things onto their existing distribution system. They need to know whether (for the same sound level) does it consume about the same amount of electrical power as the old one, for load calculations (i.e. comparison of efficiency).

2) they need to know what level sent to the distro amp (for each tap available on the step-down tx) results in the same  sound output (i.e. comparison of sensitivity).

Benta : yes, in making these measurements, of course I adjust volume for a given level on a sound meter 1 metre in front of the speaker. Otherwise it would all be meaningless. Sorry if I did not point that out.

[b_force]

If that's the case, than the approach is somewhat unconventional.

I need to know some more details about the whole system to give a more exact answer.
The things they are looking for are very easily being answered from the specs of the loudspeaker units.
That being said, you don't really need to measure the whole system if you have specs from both speakers.

If you don't, it's a very simple exercise.
First make sure you have a known output voltage. Take 2.83V for example (or a little less).
Than measure the frequency response of the system.
Don't change anything (aka, keep the same output voltage/gain levels) and measure the other system.

The difference is directly proportional to the sensitivity and the sensitivity is directly proportional to the efficiency.
The only problem is that you have the sensitivity as function of voltage.
Standard this can be calculated to Watts to use the DC resistance (Re) in the equation.

Of course this is not the real picture, but gives you a very good idea.
Another method is measuring the impedance and the frequency response of the system (amplifier and maybe even the transformer).
With that you can calculate the real power (Voutput²/Re) and therefore you know the total power consumption out of your amp.

If you have all specs and T/S parameters I would simply use a simulation program to calculate that.

[danmcb]

It's not just the LS and it's parameters. It's a 100V (distribution) paging system, where audio is distributed at high voltage (the amp has a pretty chunky step up transformer) and then stepped down at each loudspeaker unit. One amp rated nominally at 150-300W might drive anything between 5 and 50 (or so, there is no fixed number  and it depends on transformer tappings) ls units. This is a very standard technique in large installations (these days tending to be supplanted by ethernet based things, but that is another story).

It's also not for powerful PA use - output power at each speaker is low-ish (they can in fact tweak the level by changing transformer taps at each unit, but they do have to make sure that there is enough LS capacity in hand for that).

So you are looking not only at the LS and its enclosure, but also the transformer and whatever it is doing. And what you need to know is what that whole combination behaves like - in terms of Vrms in : dBa out and also W. Of course there are nominal power levels on transformer tappings, but they are fairly meaningless and  need to be backed up by typical measurements.

The main reason for the post was that I was having a bit of a brain fart about multiplying two true RMS pink noise measurements to get a stab at actual power, but then I realised I was fooling myself - all too easy to do ;-)

[b_force]

Yes, I was aware these systems (worked on them for a while).
Still means that you have to measure the system response and impedance.
No difference.