Frequency graphs of speakers....

[chris661]

Yeah it is. When the voice coil gets hot, all kinds of stuff changes.

Well, yes. Nobody said anything different.

That's quite different, though, to arguing that the response will change going from 0.1w to 0.0000000000001w. Point to the non-linear part of the speaker at those power levels. Suspension, surround, voicecoil? Something else?

Klippel testing is something you need to look at. BL(x), Cms(x), etc etc are all very linear when x is small.

Chris

 

[IG81]

I stand by this belief until I am presented by research that shows me to be wrong. Produce the research, not authority.

You chose to ignore post #16

As well as post #30 as per the above.

I think a lot of evidence has been submitted for driver behaviour, both in and out of pistonic band, within a range of operation comparable to or greater than typical usage.

 

[DonVK]

I continue to believe that extrapolation to a linear response of speakers based on one measure at 1 watt/1 meter is not science. There is no scientific expectation that we are dealing with a linear system since there is no research to show that we are. There is research to show that the frequency response changes as the volume to the speaker increases and the voice coil heats up, and the research on that has shown that its not a linear relationship.

I never said the FR changes with each tiny little change in power to the speakers. I said it wasn’t a linear change, and that we needed multiple measures at different power levels with different speakers.

I do not have the funds, the equipment, or the knowledge to do this.

I stand by this belief until I am presented by research that shows me to be wrong. Produce the research, not authority.

It's been studied alot, have a look at Kiippel's life work Measurement Overview

So far we looked at the low power end, going -40db ref 1w input. It was an interesting question and I posted some data for one driver in post #16.

Lets go the other way, the high end of the power range. This particular speaker (pc83-8) has a Pmax of around 30w. I'm willing to post some graphs on this. Lets see what i get, as speaker SPL rises by 3db for each doubling in power, and in a LA it rises by 3db for each doubling of drivers.

Single driver, power increases:
1w -> 85db
2w -> 88db
4w -> 91db
8w -> 94db
16w -> 97db
32w -> 100db

So I can safely measure distortion up to +15db ref 1W , for a single speaker.

In a LA, assuming no amp limits :
1sp -> 85db
2sp - 88 db
4sp -> 91db
8sp -> 94db
16sp ->97 db
32sp ->100db

So in an LA we could get +15db from increases in power and +15db increase from #speakers giving us 85db+15db+15db = +115db SPL max estimate. Normal loud is 90db, crazy loud is +115db so Pmax should never be a problem in LA, in a normal domestic room.

 

[DonVK]

... distortion increases at the extremes (for pc83-8).

I've never measured this before, as I seldom take a driver to/over Pmax. They just sound bad when they are really pushed. The calibration check was 85db at 1w@1m, then the mic was moved in to 0.3m to reduce room effects in the distortion measurements. All level adjustments were made at the amp to boost the signal.

It's what you'd expect, things get bad at the extremes. However it behaves pretty good up to +10db ref 1w then starts to fall apart as you approach +15db. At low frequency the Xmax (2mm) is being exceeded (and by a lot, I est +/-5mm). These are short tests so heating should not be a factor.

 

[Zarathu]

This is very helpful, and makes it easier to understand. I stand corrected Like I said I would be.

Now for another question: Since the frequency response graph of the same speaker varies from speaker to speaker(by a little amount or a big amount depending on the quality of the speaker), would it be possible to say that in a line array of of 40 speakers that it produces as a group a kind of smoothing of the whole response, much like a 1/24 smoothing? And if it doesn’t do this, what does it do, strictly from the standpoint of listening?

Is it possible to measure something like this by moving the measurement out from 1 meter to maybe 3 meters?

 

[Zarathu]

... distortion increases (at least for the pc83-8)

The FR shape seems to stay the same but the distortion increases. The levels were adjusted at the amp and the soundcard, and the mic needed to be placed at 0.1m to get a reasonable signal at -40db from the 1W reference. The increasing distortion is trending at intermediate stages as well.

However, in a line array, this is the opposite of what you hear. The distortion levels appear to collapse as you increase the number of speakers. We recognize loudness in speakers largely by the amounts of distortion. In a properly built line array, its often difficult to perceive how loud the speaker is playing: I know this from experience.

So what do you attribute that experience to?

 

[DonVK]

This is very helpful, and makes it easier to understand. I stand corrected Like I said I would be.

Now for another question: Since the frequency response graph of the same speaker varies from speaker to speaker(by a little amount or a big amount depending on the quality of the speaker), would it be possible to say that in a line array of of 40 speakers that it produces as a group a kind of smoothing of the whole response, much like a 1/24 smoothing? And if it doesn’t do this, what does it do, strictly from the standpoint of listening?

This is really a question to ask someone in the LA crowd, as this thread is for speaker (driver) FR. I think there are 3 effects. The first is the distance from the listener to the individual driver in the LA is slightly different for each driver and this could smooth things. The second is the diffraction caused from each driver center to the LA front edge (continuously increasing) causing a diffraction smoothing. Some will call it "smear" others call it "smooth". The last is the shape of the field generated be an LA that projects into the room more uniformly (cylinder).

Is it possible to measure something like this by moving the measurement out from 1 meter to maybe 3 meters?

Interesting question, does the distortion change with distance? The problem (for me right now) is its too cold to measure outside, and as I move away from the speaker I will no longer be able to gate out the room reflections.

 

[DonVK]

However, in a line array, this is the opposite of what you hear. The distortion levels appear to collapse as you increase the number of speakers. We recognize loudness in speakers largely by the amounts of distortion. In a properly built line array, its often difficult to perceive how loud the speaker is playing: I know this from experience.

So what do you attribute that experience to?

In general you will always see people trying to get the lowest noise, or lowest distortion or some improvement without asking will it matter? Often its a personal goal, and sometimes a contest.

There is another thread that is current but very long and often degenerates Who makes the lowest distortion speaker drivers that is discussing driver distortion. So far the concensus (or lack of proof) seems to be that no one can point to the exact pyschoacoustic effect (hearing impact) of a distortion graph. Lower distortion is always preferred but that is a generic observation. What distortion do we really hear or are sensitive to, is a very good question. It seems odd to me at least, that we strive for -100db in amplifiers and accept -30db in speakers.

 

[Gnobuddy]

... distortion increases (at least for the pc83-8)

Agreed!

Years ago I worked on a motional-feedback woofer system. I measured the same thing you did. I was using an accelerometer on the voice coil itself, rather than a microphone at the usual 1 metre distance.

I had mounted a small piezo accelerometer on the dustcap end of the woofer voice coil to provide the feedback signal for the motional feedback system.

During development of the servo electronics, I did many frequency sweep measurements at different power levels, looking at the output of the accelerometer, which has a one-to-one correspondence to the speakers far-field anechoic SPL.

And I saw the same thing as DonVk: negligible changes in frequency response with power over a wide range of power levels. However, distortion did rise with power, particularly at the low frequency end of the spectrum.

The most dramatic rise in distortion with power occurred at frequencies below the speakers fundamental resonance frequency. Not too surprising since that is the range where the very nonlinear spider/surround "spring" controls cone motion.

All that data belonged to my then-employer and was under a non-disclosure agreement (expired now), so I haven't seen any of it for twenty years. But I remember the trend well.

-Gnobuddy

 

[Gnobuddy]

It seems odd to me at least, that we strive for -100db in amplifiers and accept -30db in speakers.

Indeed. This is why I stopped caring about Hi-Fi amplifiers, most of which have had all distortions below audibility for decades now. As long as you keep them out of clipping, none of them have any audible "sound" of their own. Which is as it should be, because it tells you, once again, that all distortions are below the threshold of perception.

So these days I get my audio amps from the thrift store, in the form of decades-old stereo receivers that nobody wants any more. As long as they're not actually defective, they are all audibly perfect, 100% transparent, no sound of their own.

Speakers, on the other hand, are a very different kettle of fish. Every one of them is coloured. I have never heard two that sound identical, at any price point. Which tells you that each one has defects that are more than large enough to be audible, and that's before we even start considering the interaction with the room they're in.

-Gnobuddy

 

[weltersys]

1)The distortion levels appear to collapse as you increase the number of speakers. We recognize loudness in speakers largely by the amounts of distortion. In a properly built line array, its often difficult to perceive how loud the speaker is playing: I know this from experience.
2)Now for another question: Since the frequency response graph of the same speaker varies from speaker to speaker(by a little amount or a big amount depending on the quality of the speaker), would it be possible to say that in a line array of of 40 speakers that it produces as a group a kind of smoothing of the whole response, much like a 1/24 smoothing?
3)And if it doesn’t do this, what does it do, strictly from the standpoint of listening?
4)Is it possible to measure something like this by moving the measurement out from 1 meter to maybe 3 meters?

1)Since multiple drivers are used in a line array, excursion is far less per driver at the same level of low frequency reproduction, so harmonic distortion is also less at a given level. A small single driver can't reproduce rock concert levels at more than a few inches before gross distortion is an issue, given enough drivers, harmonic distortion becomes a non-issue, replaced by many other issues.
2) Yes, the response of a line array in the far field is an average of the individual driver's response, and does look like "smoothing" has been applied.
3) A simple line array's response is dictated by the listening position.
4) One of the most useful "features" of a line array compared to a single speaker is the perceived difference in the inverse distance loss- ordinarily 6 dB per doubling of distance, a line array has around only a 3dB loss per doubling of distance in the transition from near to far field.
This can be very useful to even out distance variations in level, but the reason it occurs is high frequency (HF) summation in the far field, and HF path length interference patterns (comb filtering) in the near field. All this can be measured at various distances to "get a feel" for why a line sounds like it does, though it really is easier to hear what the "line" does outdoors, room response complicates measured and perceived response immensely.
The interference patterns in the near field reduce HF SPL near the line compared to further away, and are also responsible for the narrowing measured vertical dispersion- off axis response is cancelled depending on frequency and line length.
In the near field, the lower frequencies sum, the higher frequencies sum at progressively greater distance, hence the line array, having a higher direct to reflected HF sound ratio "sounds clearer" further away than a single driver's wider vertical response exciting more floor and ceiling reverberation.
At the same time, a single driver would "sound clearer" up close because of the lack of multiple arrival time differences the line introduces.

Every design has compromises: single full range drivers, multiple driver virtual single point sources, multiple full range driver line arrays, multi-way, etc.
I use all five of the above designs around the house and for sound reinforcement work.

The difficult (or fun) part is choosing which design is best for the environment it will be used in, and it's component and build cost compared to the listening enjoyment provided.

Cheers,
Art

 

[DonVK]

However, in a line array, this is the opposite of what you hear. The distortion levels appear to collapse as you increase the number of speakers. We recognize loudness in speakers largely by the amounts of distortion. In a properly built line array, its often difficult to perceive how loud the speaker is playing: I know this from experience.

So what do you attribute that experience to?

I missed my chance to add this to the original reply, since it occurred to me afterwards.

Let's work backwards from say needing 91db to compare the effect on a single driver.

Our single driver (85db/w@1m) on its own will need +6db ref 1w (or total 4W) to hit 91db and may still drive clean. In the LA each driver is at 91-15db=76db or -9db ref 1w and that's 0.125W of power per driver. So its more likely the driver will behave better with 0.125w compared to 4w. The previous graphs looked at a single driver from -40db to +15db ref 1w.

 

[DonVK]

Agreed!

Years ago I worked on a motional-feedback woofer system. I measured the same thing you did. I was using an accelerometer on the voice coil itself, rather than a microphone at the usual 1 metre distance.

-Gnobuddy

I have an interest in this subject as well. In amplifiers (as all systems), the performance improves with closed loop feedback. Speakers still run open loop with possibly a static feedback loop during EQ measurement to correct the FR response.

Accelerometers are certainly one measurement method, after all music is AC, so velocity and acceleration are all derivatives of position. I have been considering a laser reflected off a contrast strip on the back of the driver that could provide relative position data to "close the loop". We know the input signal, we can model the driver + enclosure, so we just need it's position.

 

[Gnobuddy]

...HF path length interference patterns (comb filtering) in the near field.

And I suspect this comb filtering is the major reason for the "shrieky" treble that line arrays tend to produce.

I have had the misfortune to hear both the Fishman SA220 and the Bose L1 a number of times. Both suffer from unpleasantly "shrieky" treble, and I found this extended all the way to the rear and side walls of the venues; if the near-field treble comb filtering is supposed to disappear in the far field, there was no audible evidence that this was in fact the case.

Incidentally, the very popular Shure SM58 microphone has a ragged high-frequency response with multiple deep notches and peaks that seem to be caused by mechanical break-up modes in its diaphragm. It also has a similarly "shrieky" sounding treble.

The same problem afflicts those piezo horn tweeters that surfaced a couple of decades ago, and still turn up from time to time. They sound shrieky and harsh, and if you look up published frequency responses, they are ragged and full of deep valleys and tall peaks.

I don't know why the human ear/brain listening mechanism hears comb filtering / multiple deep notches and peaks as sounding "shrieky", but that does seem to be the case.

Then again, both the Shure SM58 and Bose L1 have many satisfied customers, so not everyone is bothered by the sound of a ragged high-frequency response.

-Gnobuddy

 

[wesayso]

I haven't heard the Bose L1, but I do know the character of the driver used is amplified in a line array. Which is why I would advise to not just take any driver and try it.
More than a few tried and were dissapointed.
Our ears hear comb filtering all the time, no way around that for people with two functional ears, so you'd have to find some other factor to put the blame on.

 

[Gnobuddy]

I have an interest in this subject as well. In amplifiers (as all systems), the performance improves with closed loop feedback.

I managed to get the servo stable with 18 dB of feedback at midband. Measured 3rd harmonic distortion dropped by almost the same amount, 15 to 18 dB, at some frequencies and power levels.

It wasn't all "Happy, happy, joy, joy!", however. I discovered that many commercial recordings had bass that sounded too dry and toned-down when played back through an accurate woofer with motional feedback. Presumably they were mixed to sound good on speakers with some "woofiness" in the bass.

Accelerometers are certainly one measurement method, after all music is AC, so velocity and acceleration are all derivatives of position.

As you say, position, velocity, and acceleration can all be used to supply the signal for motional feedback. However: the near-field air pressure is proportional to cone velocity, not position (as you'd expect from the Bernoulli equation).

And when it comes to woofers, whose diameter is much smaller than the wavelengths of the sounds they're radiating, there is an additional frequency dependency: the efficiency of radiative coupling to the air is proportional to frequency, meaning the SPL at a normal 1 metre distance is proportional, not to the cone velocity, but to the cone acceleration.

This is somewhat of a coincidence. It happens because for sinusoidal motion, multiplying the velocity by the frequency in fact gives you the acceleration...(to within +/- the constant 2 pi, depending on whether you like your imaginary exponent in the complex angular frequency positive or negative.)

So: there are good arguments to be made for starting out with a sensor that responds directly to either voice coil velocity, or voice coil acceleration.

Basically, if you use feedback to flatten the frequency response of the acceleration, you also flatten the frequency response of the sound emitted by the speaker, which is what we are actually after!

If you use a sensor that directly responds to voice coil position, on the other hand, (and not velocity or acceleration), you may find you have to build a differentiator into your servo loop, and they tend to be noisy and unstable.

I have been considering a laser reflected off a contrast strip on the back of the driver that could provide relative position data

My initial thoughts were similar, in fact I prototyped a sensor that moved the angled edge of an X-acto knife blade between an LED and photodiode, so that variations in voice coil position produced corresponding variations in photocurrent.

On reflection, I realized that position was probably not the optimal signal to derive. As a bonus, it was a lot easier to get a good clean acceleration signal, than to make a vibration-free LED/photodiode mount!

-Gnobuddy

 

[Gnobuddy]

I do know the character of the driver used is amplified in a line array.

As it happens, I recently had a conversation with one of my musician friends about the Bose L1. He said it sounded like a small driver being thrashed hard...only louder.

Many people seem to believe that a line array will sound better than the individual drivers used. Subjectively, the opposite seems to be true; my experience is that small 2" or 3" drivers that sound just mediocre individually in a TV, sound outright bad when strung into a line array. This seems to agree with what you just wrote.

so you'd have to find some other factor to put the blame on.

Maybe so. Nevertheless, sound sources with jagged frequency responses sound shrieky to me.

Perhaps it has to do with the nature of phase shifts or transient ringing of nearby "lobes" in the comb filter. Maybe it's not comb filtering caused by acoustic path length differences that causes the shrieky sound, maybe it's because of the multiple not-quite-equal cone breakup frequencies of the various drivers in the line array, with accompanying unrelated and time-varying phase shifts of each ringing breakup mode. I don't know.

What I do know is that comb filtering caused simply by direct line-of-sight path differences to each ear, produces a predictable and stable phase shift as frequency or listening position changes. And that doesn't cause that shrieky sound.

Listening to two guitar speakers fed the same electrical signal and placed 10 feet (sorry, 3 metres) apart doesn't produce that shrieky sound, either.

But string 50 P.A. horns in a grid across a football field or a fairground, feed them all the same signal, and it does sound shrieky if you happen to be in the line-of-fire of several of them at the same time. I remember having this experience at some low-quality outdoor music events. To be fair, it's hard to say if one driver by itself would also have sounded shrieky: it was impossible to hear one by itself.

The same thing seems to be true for a string of small drivers in a long skinny tube enclosure, a la the Fishman SA220 or Bose L1. They sound shrieky, for whatever reason.

-Gnobuddy

 

[DonVK]

I haven't heard the Bose L1, but I do know the character of the driver used is amplified in a line array. Which is why I would advise to not just take any driver and try it.
More than a few tried and were dissapointed.
Our ears hear comb filtering all the time, no way around that for people with two functional ears, so you'd have to find some other factor to put the blame on.

What was the basis for your driver selection ? was it extrapolation from the single driver ?

 

[wesayso]

Somewhat like this:
First and above all else: clean IR/waterfall.
Preferably a rising frequency response on axis (only for array duty)
Good off axis performance out to 15 degree, mimicking on axis.
Distortion

Validated in listening tests as a mono Fast with 15"woofer (using DSP as I don't like a rising response for a single driver to listen to)

Choosing between TG9 and TC9 was tough. I went with paper due to having preferred paper in previous setups.

 

[DonVK]

Given that you eventually need to buy 50 or so, its a small "insurance price" to pay, to just sample test a couple driver models first

Even with those criteria, you still need a far amount of judgement to decide on a driver from the sample set. Off axis +/15 deg should be possible for most smallish drivers. The IR and distortion are usually not supplied by the mfg so you have your own setup / eq / measurements. Then there is still the issue of setting the acceptance thresholds and what sound you like. Or just using what has worked before.

Did you buy extra drivers (50+) and screen test them to get a consistent set with lower variance?