Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[iain]

Hi

I’m an acoustics student doing a bit of uni course work on loud speakers at the moment.

I’m trying to a design a set of criteria that accurately measure every aspect of loud speaker performance.

So far the test is

        ============================================================

Measurements 1-8 should be made at 101cm and 233 cm.

Frequency Response

1.   Frequency range - On axis pure tone sweep to establish lower and upper -6dB point, measured in free field.

2.   Frequency variations - pure tone sweep across frequency range measured in 1, in free field, at vertical and horizontal intervals of 11 degrees from 0-352 degrees and 8 degrees from 352-360 degrees.

Dynamic Response

3.   Dynamic range - free field maximum sustained SPL in dB at 1 meter on axis with 10 minute of continues pink noise ascertaining speakers peak output level.  

4.   Transient response - using standardized impulse signal at 0.1w 1w and 10% of peak measured in free field at vertical and horizontal intervals of 11 degrees from 0-352 degrees and 8 degrees from 352 - 360 degrees.


Timing Response

5.   Frequency and amplitude dependant timing variations - measured measured in free field at 1/3 octave bands at 0.1w 1w and 10% of peak between speakers upper and lower - 6db frequency points, at vertical and horizontal intervals of 11 degrees from 0-352 degrees and 8 degrees from 352 - 360 degrees.
   

Harmonic Distortion

6.Measured distortion - between speakers upper and lower -6dB frequency points, measured in free field at 0.1w 1w and 10% of peak at vertical and horizontal intervals of 11 degrees from 0-352 degrees and 8 degrees from 352 - 360 degrees.

7.   Results can be deduced from 1, 2, 4, 5 and 6.  


Amplifier Requirement Specification
 
8.   Sensitivity - at 0.1w 1w and 10% of peak (using standard method)

9.   Impedance, inductance and capacitance sweeps - measurements made between speakers upper and lower  -6db frequency points at 0.1w 1w and 10% of peak.
 

        ============================================================

That’s about everything I can think of.

My question is whether or not people agree with this statement and if not for what reason.  

“This test will quantify every perceivable audible difference in loudspeaker performance.”

Cheer

Iain  

[franman]

Ian,

You seem to have a thourough set of tests, BUT to make that statement, that is is possible to quantify ever perceivable audible difference... that's a little bit bold. I'm going to ask Lars to have a look at your battery of tests.. He's the expert on this, but I don't believe it is possible to quantify every perceivable difference between loudspeakers. We've done A LOT of measuring and A LOT of listening over the years... we hear a lot of stuff that's difficult to measure, not only with loudspeakers but with converters, analog signal paths... everything.

I guess I just believe that the ears, along with the brain, are the best test instrument we've come up with... Of course, it's difficult for them to remain objective, even in blind tests because we 'want' to hear a difference to justify our (audio) existance... (How's that for a bold statement??)

Lars... what do you think about these tests???

[iain]

Hi Francis

I’m not really making the statement with confidence, infact im very probably wrong,    the statement is really to just to see if people know of any other acoustic phenomenon I need to take into consideration.

With regards to the signal chain, we’ve got a program called win MLS and im told with some clever mathematics it can analyse the signal chain and subtract its effects from the measurement.

One of the reasons that I’m including this in our project is because the current system of loudspeaker measurements seems to fall drastically short of quantifying differences in mid – high end loud speakers. Attached is two spec sheets, from a company that supply’s more test specs than most than most. One speaker retails at around

[iain]

[Tom C]

iain wrote on Mon, 23 April 2007 12:48

Technical specifications say not much about a speaker's performance
for a listener.
I've build lots of speakers and very often small/tiny changes
(which don't (measurable) change the things listed in the specs)
cause dramatic changes in sound.
Things as crossovers are very simple to build but are from an
electro-engineering point of view very complex.
And speaker housing design is a science of it's own.
De-tune the pipe of a bass-reflex speaker just a little bit
and it sounds completely different.

Tom

[iain]

Hi tom

I take your point, but the days of  not having accurate enough equipment are over, our university main anechoic chamber has a background noise floor between 100Hz-20Khz of -15db they know its that low cause they’ve got a mic that can measure it, our ears certainly cant.

I appreciate that the test spec shown above don’t quantify the change of  an expensive inductor  or  an extra 10g of foam on the back wall of your speaker, however I do think these effects would show up clearly in the test I proposed in my first post. I would even say that it would show up more clearly than by subjective listening. Unless you do your listening in very controlled anechoic test condition how can you be sure that the increase in 65 Hz bass caused by what you think is a shortening of the port is not caused by a few cm change in your head position, the increased activation in room resonance caused by a slight change in speaker positioning or a drop in background noise cause by a cooling fan switching off.

Last year I put a post up regarding a

[L_Tofastrud]

Iain,

I think as a research subject this can probably be interesting but a loudspeaker is a very complex thing and unless you quantify the contribution of the individual parts (drivers, x-overs, cabinet) I'm not sure what conclusions can be drawn in a side by side comparison of two speakers.  I think you'll end up with a lot of data but how do you know what is "better"?  Is there such a thing as an idealized mathematical performance for all the measurements you described?

Do you have access to an anechoic room?  What is the low-frequency limit of this room?

Are you familiar with the KLIPPEL measurement system?

If I might:  instead of using Watts I'd use Volts (1V and 2.83V) and I would also include a measurement with a given dBSPL (small speakers 96dBSPL @ 1 m, larger speakers 96 dBSPL @ 2.5 m).  This would produce easily comparable measurements.

Regards
Lars Tofastrud

[iain]

Hi Lars

With regards to the anechoic camber, I’ve been trying to find the commission data, hopefully post it next week, but the wedges are 0.95 m long indicating a combined porous absorber length of 1.90m meaning at first glance the room is fully anechoic down to 45Hz using the ¼ wavelength assumption, however were fortunate to have Trevor Cox of RPG acoustics as one of professors and I suspect he probably used a form of resonant absorber to soak up the rest of the low frequency energy. As I say ill pester someone for the data and hopefully post it next week. I cannot book the chamber at the moment I will be able to as of September this year to research my dissertation.

With regards to the question of perfect loud speaker, and measurements, the goal must surely be for something that produces precision of accuracy in reproduction that is bellow the average human threshold of perception.

Anyway its all very well making a statement like that but what do I think it means in measurements. This is a really busy time of year for me so I dont have a lot of time to throw at it, so for the moment I’ll do freq response ill get round to the rest over summer.

I made some random number generators in matlab and off axis co-efficient, low roll off, freq variation ect. So I can put in theoretical performance criteria and generate something like the data I might get if I performed my 1, and 2 freq response measurements.

If I plot it as 180o on axis, in a 3d mesh I get graphs showing the speakers complete frequency response, there would be another one at the second distance but who knows how they would correspond, I’m not even going to try and simulate it, for now ill just do these 4. sorry the z axis is a bit wrong but you get the idea.  

Below what a crap speaker might look like, +-5dB ect  

ok_speaker2 is on axis +- 3db, Bessel function behaviour above 6 kHz, roll off at 50Hz with an off axis variation multiplier of 0.7, a fairly typical hifi loud speaker probably, If I ever do the test I expect to see results similar to this.

good_speaker2 is on axis +- 1 dB Bessel function behaviour above 15 kHz, roll off at 30Hz with an off axis variation multiplier of 0.3. I recon even the best loud speakers would struggle to better this,

perfect_speaker2 this is the point of theoretical inaudibility, all frequency variations at all points less that +-0.5dB with in the audible range. A this point I would suggest further improvement would be futile as most humans can’t hear it.

I don’t know if perfect_speaker2 is possible but I think it should be a goal in performance, but without testing how will we ever know.

I hadn’t heard of KLIPPEL but it does look fairly comprehensive, perhaps the loud speaker industry should use it to standardize some more comprehensive testing.    
 

[iain]

[iain]

[iain]

[Tom C]

Okay, you've got now a speaker with a perfectly flat response
curve for a constant sound source.
That's the frequency domain. Fine.

But there's also the time domain with impulse response.
You can measure this, no problem.

But you need mechanical devices (at least at lower frequencies)
to move the air.
And these create distortions in the fequency domain with each
impulse.

Now the question: which of the 2 domains do you want to compromise more?
Which compromise sounds best? Objectively? Subjectively for
the human ears?

Speakers haven't changed much over the last couple of decades, but
still no one can predict how a given new speaker design will sound
in reality.
Can of worms.

Tom

[franman]

major can or worms....

[johnR]

A few more characteristics I can think of: intermodulation distortion, Barkhausen noise and power compression. Those first two are almost never mentioned, but have a huge effect on a speaker's resolution.