How important is measuring speakers?

[Spanky Ham]

I have always thought that measuring speakers with equipment in their enclosures was the only way to design a decent speaker. I haven't given much thought to this until recently when I was over at PE (a place I rarely visit). I know there are a couple of guys who are designing speakers without equipment with one guy who seems to be posting a lot on a couple of forums. What got me curious was one notable poster (I think) over at PE designs his speakers without equipment. I believe his speaker designs have received positive feedback. Is designing speakers like color matching in my profession? I use a spectrophotometer to always get a delta error of around .5 for color matches. Someone doing it by eye may get close and may even stumble around and achieve a .5 de, but most will be between 1 and 2 de. Most people can see a 1 de difference and no one would miss a 2 de. I am not sure if my analogy works or not. So, my question is can one design a decent or average speaker without equipment?

[cjd]

When one knows enough and makes all the right assumptions, yup.

As I said in another thread here, I often sketch that way. I've improved dramatically in my "sketches" over time as I learn where my assumptions fail to meet reality. But working with measurements always land a better result. Can you get a decent result? I think so. But it's unlikely that it will be as good as you could do with real measurements (in part because real data means you can look at far more data points, different axis measurements, different output levels...) Could the average person hear the difference? Probably. Depends in part on the designer's experience working BOTH worlds. How much z-axis offset to use. What box modeling might miss. Lots of stuff.

It would be like having someone that's only learned a bit of theory do color matching without measuring and without actually testing the mix.

The question then becomes, how close can you get? I'd say, that's all down to luck. There are far too many cases where a design in this manner fails to work properly (I have as many failures as basic successes under my belt). And as many cases where a designer that knows what their doing puts out a design that "fails" this process. The process is where the shortcoming exists, because it only works within a limited scope.

I've got a design or two in the sketch portion of my process right now. Tempted as I am to buy all the crossover parts and build away... I'll be patient and wait till I've got real data. And I'll be quite curious to see how close I got. And how far away I really was.

I'm not sure what area you work, but I always have fun when it comes to mixing paint pigments. My training is in fine art, and I learned a very different approach - I never had any white or black to work with. A bright yellow, a red-orange, a red, a lighter blue, and indigo. I think that was it. 5 raw pigments. Though in the back of my mind I have this nagging thought that it was 6. And I never failed to succeed in painting the right color.

[Andy_G]

deleted

[TacoD]

Imho you need measurement equipment if you want to correct driver anomalies and use steeper x-overs. Also you need the actual measured performance of the driver in the cabinet, there ares some tools which can predict this, but I want to see the FR. If you have measured and made a lot of filters then your estimated guess can come very close, but you would be surprised how the FR looks of a loudspeakers which are tuned by ear.

[dlneubec]

I'm no expert, but I suspect that the success rate for designs without measurements is directly related to how complex the design is. A simpler two way with forgiving drivers would likely be the easiest. Throw in metal cones and it gets more difficult. Make it a 3-way and it takes another step toward it being unlikely you will be successful. Now throw in an unusual driver arrangement and your success rate drops even further.

My experience with omnidirectional speakers suggests to me that there would be no way to do a good one without measurments, since simulation software is not set up to deal with anything this unusual, and the published data on driver FR is unuseable. I'm sure there are other cases like this where the software is just not available to do a decent simulation or the manufacturers data would be unusable.

[---k---]

Have you ever noticed how many people think a HTIB is wonderful? Or how glowing the reviews are for the first DIY design someone has built, even though it is a very budget or compromised design? And lets not get sidetracked into Stereophile's reviews. Anything can look pretty impressive to someone - especially on the internet.

I got the impression the comment was referring to someone else at PE. I think you have to take everything you read on the internet with a healthy bit of skepticism.

That said,

I've heard some designs at Iowa that were done without measurements. I think you can get decent to good results without measurement. So, I guess it depends on what your goals are. If you're just screwing around, w/o measurements are fine. If you want to design a world beater, you better have measurements.

[crazybastard]

I agree with CJD that you can make a reasonable attempt at designing speakers without major measurements if you stay within a reasonable limit of audio physics.
20-20khz flat is the ideal theoretical speaker and is the ideal for every design. IMHO it's the imperfections from this ideal that give the speaker a certain sonic signature.

[fjhuerta]

The only time I've been able to make any assumptions were while building a 2 way speaker with a wide range Altec woofer and a supertweeter. The woofer was running full-range (it was that smooth!) and we only added resistors and caps to the tweeter. It was a surprisingly nice sounding speaker (it had more to do with the woofer than the crossover or the supertweeter).

But I wouldn't try anything more complex... just the thought of diffraction makes me scared.

[ThomasW]

Without measurements there's no way to know how much unit to unit variation exists between drivers. Assuming the drivers you buy today will have identical performance characteristics to ones measured a year ago is problematic.

Our thread discussing the variations in the D26 tweeter is proof positive of why the idea of using only someone elses old measurements and software sims is a fools errand...

[Kevin Haskins]

Without measurements there's no way to know how much unit to unit variation exists between drivers. Assuming the drivers you buy today will have identical performance characteristics to ones measured a year ago is problematic.

Our thread discussing the variations in the D26 tweeter is proof positive of why the idea of using only someone elses old measurements and software sims is a fools errand...

I totally agree.... Simulations are great but they rely on good data. Garbage in, garbage out as they say.

[Spanky Ham]

Thanks for the feedback.

Ryan,
I am glad you were at Iowa. I hope I am not stepping on any toes, but the person that got me thinking about this is Pete Shumacher. Did you find his design good or could it have been better?

[Maximiliano]

Without measurements there's no way to know how much unit to unit variation exists between drivers. Assuming the drivers you buy today will have identical performance characteristics to ones measured a year ago is problematic.

Our thread discussing the variations in the D26 tweeter is proof positive of why the idea of using only someone elses old measurements and software sims is a fools errand...

We should note that sampling variation of a driver is a little different matter here. If that is a major reason for measurement, building any of the projects in the Mission Accomplished will also be a fool's errand. This means that one should measure each driver and adjust the crossover accordingly.

But this is usually not the case. A careful designer reports any version change of drivers used in his design. Zaph is very sensitive to this matter and doesn't forget to add a note in his measurement pages.

Max

Edit by moderator to add
Maximiliano is Woe Jay Kim pretending to be someone else

[Maximiliano]

Thanks for the feedback.

Ryan,
I am glad you were at Iowa. I hope I am not stepping on any toes, but the person that got me thinking about this is Pete Shumacher. Did you find his design good or could it have been better?

I know him and some other people (e.g., Wolf) who build speakers without measurement at the PE board. An important thing you need to know about the practice they do is that their actual build is not simply based on one shot simulation. They have accumulated skills and knowledge from building a lot of speakers that way. They know how to use simulation tools not only for sketching a design but also for modifying and voicing it. Also, from experience they usually know how to find a problem by listening and how to fix it in a crossover.

But I agree with Dan that a complex design requires measurement. No way to design his omni-directional speaker without measurement.

Max

Edit by moderator to add
Maximiliano is Woe Jay Kim pretending to be someone else

[jkrutke]

When I started posting driver measurements, I never intended for people to be tracing them to simulate designs. The real intention was to help designers make a purchasing decision. That's why I never posted FRD files or SE driver files.

There are many things that have to be guessed at. Acoustic centers, real off axis behavior rather than simulated based on diameter, system harmonic distortion performance with the crossover in place, and on and on. Without measurements, all that can be done is guessing based on what other people have done.

People can sim designs all they want, build it and be happy with it, others can build it and be happy with it, but the lack of measured data will always hang over the design like a dark cloud. Still, it's a fun way for a casual speaker designer to play around and learn things, and at least doing some simulation is still a big step above just using ears and a pile of crossover parts.

[Jim Holtz]

Thanks for the feedback.

Ryan,
I am glad you were at Iowa. I hope I am not stepping on any toes, but the person that got me thinking about this is Pete Shumacher. Did you find his design good or could it have been better?

Pete has brought speakers based on his sims for the last two years. Each year he has taken them home and changed the crossover based on the feedback and this year based on measurements. Pete does get them pretty close though. Many people would be happy as is if they didn't have a chance to compare them to others. IMHO, measuring equipment should be the first thing for a crossover designer to invest in.

Sims are the beginning, not the end. Lots of measuring and voicing is required in between.

Jim

[---k---]

Thanks for the feedback.

Ryan,
I am glad you were at Iowa. I hope I am not stepping on any toes, but the person that got me thinking about this is Pete Shumacher. Did you find his design good or could it have been better?

Pete's speakers were good. Enjoyable to listen to. Beyond that, I have some opinions, but I don't think that I'm experienced enough to give public critiques. I think that should be left to those that have designed a few speakers themselves. Don't take that to mean I think there were major flaws - just fact that I don't publicly bash people's designs without a lot of factual basis. Also, I'm don't really feel that the 10 minutes of time in Iowa gave enough time to fully judge.

Oh and finally, they were in the Midrage class, and I'm used to an Unlimited class speaker.

[Paul Ebert]

Should one have the necessary hardware, how effective would it be to use a tool like Arta for the measurements?

Also, do most of you use crossover simulation to do your tweaking or do you physically swap parts in and out?

[cjd]

I use software and parts swapping.

[Dave Bullet]

I am beginning to think the ability to measure harmonic distortion is critical.

I have measured, designed a passive xo, implemented then re-measured. What I got was the intended on-axis and off-axis response. (Flat on axis with no serious off-axis issues until 60 degrees then tweeter directivity beyond 7KHz).

What did it sound like? - very forward / bright (aka crap).

Either it was the woofer crossed over too high and upper midrange harmonic distortion (the speaker was very loud in the upper midrange) or poor BSC choice.

I appreciate Zaph's data for evaluating drivers for selection and concur than your own measurements must be done in your own environment.

I just wish manufacturers would post harmonic distortion data data too.

Cheers,
David.

["2K"]

I have measured, designed a passive xo, implemented then re-measured. What I got was the intended on-axis and off-axis response. (Flat on axis with no serious off-axis issues until 60 degrees then tweeter directivity beyond 7KHz).

What did it sound like? - very forward / bright (aka crap).

Cheers,
David.

Given the fact that the low end is difficult to measure accurately with the short gate times most people use BSC usually needs to be finalized by ear.

Is your mic calibrated?

[Dave Bullet]

Given the fact that the low end is difficult to measure accurately with the short gate times most people use BSC usually needs to be finalized by ear.

Is your mic calibrated?

Hi 2K,

No it is not. I tried to get it calibrated locally (NZ). I know Kim Giardin (?) did this in the states, but I can't justify the cost of return postage to have this done.

Mine is the common Behringer ECM8000. Of course this will change between production runs and model of Panasonic caplet used.

I did a crude form of self calibration. I used a Seas tweeter which gets measured pretty consistently, then measured against Zaph's results, then found a calibration file on the net that closely approximated Zaph's results. Generally, the ECM8000 is good until 2KHz from which generally Kim said there is a rise until a resonant mode in the 12 - 15KHz area where a peak occurs.

My current "reference" speakers are Kef-C30s. They sound balanced and "right" to my ears. When I measure these, they are flat apart from a little rise around 9Khz and again a bump around 60Hz (likely to be a room mode). I therefore feel my crude calibration is sufficient.

My DIY attempts measure very similarly to the Kefs (same measurement conditions) but sound more forward. Bass is about the same.

My next step is to measure HD of both the Kef's and my DIY to see if that highlights the problem.

Thanks,
David.

["2K"]

My current "reference" speakers are Kef-C30s. They sound balanced and "right" to my ears. When I measure these, they are flat apart from a little rise around 9Khz and again a bump around 60Hz (likely to be a room mode). I therefore feel my crude calibration is sufficient.

My DIY attempts measure very similarly to the Kefs (same measurement conditions) but sound more forward. Bass is about the same.

My next step is to measure HD of both the Kef's and my DIY to see if that highlights the problem.

Thanks,
David.

Hi Dave.

Fair enough about the calibration.

What kind of measurement did you take that shows 60 Hz in room with sufficiant detail to make a comparison with another speaker? I'd guess you would have a lot of roughness in the FR plot from reflections.

You could try adding 1 - 1 1/2 Db of BSC and see what happens.

[Dave Bullet]

Hi Dave.

What kind of measurement did you take that shows 60 Hz in room with sufficiant detail to make a comparison with another speaker? I'd guess you would have a lot of roughness in the FR plot from reflections.

Hi 2K,

I did a farfield ungated measurement with microphone at listening position with both speakers in their normal position (isosceles triangle) playing a mono white noise sweep. (After all - ultimately I'm really interested in what I hear where I normally sit - not at 1m on tweeter axis).

I suspect a room mode as my living room is 5.7m by 7.5m with a cathedral ceiling. 5.7m roughly equates to 60Hz (one wavelength), but I always forget whether room standing waves are excited at the 1/2 or full wavelength.

I could post it here... but I am really taking this thread off topic so won't.

Hi Dave.

You could try adding 1 - 1 1/2 Db of BSC and see what happens.

I changed tweeter padding to create a dip around the XO point (2KHz in my case) and that made the speaker much more listenable. Some would say it is the tweeter at fault. In this case it was a 27TDFC/G (the grille version of the 27TDFC) so I don't think so. I still think there are harmonic distortion products on the upper end of the woofer. All I want from Santa is some time away from household chores to do some more speaker measurements.

cheers,
David.

[tpremo55]

Dave,
Is it possible that Phase correlation between the drivers has anything to do with the sound you are describing (very forward/bright)?

["2K"]

Hi 2K,

I did a farfield ungated measurement with microphone at listening position with both speakers in their normal position (isosceles triangle) playing a mono white noise sweep. (After all - ultimately I'm really interested in what I hear where I normally sit - not at 1m on tweeter axis).

I suspect a room mode as my living room is 5.7m by 7.5m with a cathedral ceiling. 5.7m roughly equates to 60Hz (one wavelength), but I always forget whether room standing waves are excited at the 1/2 or full wavelength.

I could post it here... but I am really taking this thread off topic so won't.



I changed tweeter padding to create a dip around the XO point (2KHz in my case) and that made the speaker much more listenable. Some would say it is the tweeter at fault. In this case it was a 27TDFC/G (the grille version of the 27TDFC) so I don't think so. I still think there are harmonic distortion products on the upper end of the woofer. All I want from Santa is some time away from household chores to do some more speaker measurements.

cheers,
David.

Hi Dave

Enjoying your summer down there? Lots of snow around here.

A white noise sweep? I am unfamiliar with this method. I will normaly do an RTA with pink noise to generate a room response.

A gated measurement can be taken at a longer distance than 1 M with a little attention to objects that can cause early reflections.

It looks to me like you have a very good sized room and reasonably well proportioned for listening. I'll bet most of us have to make do with smaller and more problematic areas.

Are we that far off topic? It is about the need to measure for good results which I consider very important. Maybe a different thread is in order to address your specific issues.

I'm glad to see you made some progress with padding the tweeter.

What woofer are you using BTW.

Happy holidays!

[cjd]

Dave,
Is it possible that Phase correlation between the drivers has anything to do with the sound you are describing (very forward/bright)?

Sure. Could be phase. A 3-way with the woofers wired out of phase will do this (ask Ryan ) Could be box stuffing. Could be driver breakup/distortion. Could be a great many things.

C

["2K"]

Sure. Could be phase. A 3-way with the woofers wired out of phase will do this (ask Ryan ) Could be box stuffing. Could be driver breakup/distortion. Could be a great many things.

C

Wouldn't a driver wired out of phase show up on an FR plot? David is claiming it to be flat or at least reasonably so. He said he made off axis measurements out to 60 degrees. I'm assuming he included vertical off axis. If so that would show phase correltion issues.

[Dave Bullet]

Hi all,

Sorry for the delayed reply - been camping with the family. 5 kids - lots of fun :B Just got back. house is safe (always worried about burglaries and my vintage amps being stolen). Anyway....

Hi Dave
Enjoying your summer down there? Lots of snow around here.

Yup - we had a warm summer. 8 days of camping and only one day of rain. High 20's low 30's most days (celcius) so quite pleasant.

A white noise sweep? I am unfamiliar with this method. I will normaly do an RTA with pink noise to generate a room response.

Speaker workshop will normally do white noise, or use pink noise when you use the "amplifier calibration" setting for measurements. I tried both and couldn't see the difference. Apparently, pink noise is better for low end resolution, but this will get swamped by boundary effects in your measurements anyway so gating is required.

What woofer are you using BTW.

Happy holidays!

Here is the design attempt here:

http://homepages.paradise.net.nz/~davebullet/speakers/Wharfedale505_2/main.html

all measurements are 1m gated with woofer nearfield spliced in.

The woofer is a Wharfedale proprietary 8" driver from the 50x series produced in the late eighties.

I already received a bit of help over at diy audio when I first did this. At listening position, there is a small 6KHz peak (not shown in the 1m gated measurements on the website above) that some suspected was the cause of apparent brightness / forwardness. I can only guess this is a product of the room, since the Seas tweeters have no known issues at that FR.

Cheers,
David.

[Spanky Ham]

I usually hate resurrecting threads, but what the hey.

I understand the voicing of speakers. I believe George Short has stated in the past that he usually does the measurements and then finalizes the design by listening to it over many months. I guess in my original analogy it would be akin to looking at a color match over a period of time and adding a drop of pigment here and there to get as close to perfect as possible.

After thinking about this for months (and I don't know why I haven't posted earlier), is the absolute best sound system the one designed for the room it is to used in? I mean if the room has a huge influence in the quality of sound, then it would stand to reason that to get the best a speaker would have to be built for that room. I believe this is what Mark Seaton does, but hopefully someone will enlighten me?

[Dave Bullet]

is the absolute best sound system the one designed for the room it is to used in?

My uneducated 2c

I would say yes I believe most of the effects / problems will be in the balance of bass to the mid and treble frequencies and placement (or constraints on placement) of speakers with respect to boundaries (side walls and back wall).

Your usual listening position relative to the speakers would also affect the bass tuning (BSC) and also probably crossover. Nearfield usually means less BSC is required. If your usual listening position is < 1 m, yet the designer has optimised the crossover for farfield listening, then crossover work may be required to optimise phase matching.

David.

[ThomasW]

Mark doesn't custom design individual speakers for every different room. For the HT installations he's posted about Mark uses DSP to optimize his standard design (the Catalyst shown below) for the acoustics of the clients room.

[Dennis H]

What Thomas said.

While most of us don't have DSP crossovers, many of us do have receivers with room EQ (fancy tone controls) that can adjust the 'voicing' of the speaker. It seems to me the main thing in designing a passive crossover for everybody is getting the blend (SPL, phase, etc.) right between the drivers and operating each driver in its low-distortion range. Then use the the receiver's EQ to adjust for how close you place it to the wall or how hard or soft your room surfaces are., i.e. the voicing in your particular room.

Edit: so yeah, measuring in your room is important, whether you prefer to do it yourself or trust your receiver to do it.

[jquin]

What's the consensus on use of HT receivers room EQ?
I thought most HiFi buffs had contempt for them.
I'm sure that using room EQ is not as good as a rectangular room of the Golden Ratio with speaker placement of Golden Ratio but lets face it this does not suite most HT installations.

I wonder what the lesser of the two evils is,

consumer DSP room EQ
or
unequalized room with non-optimal speaker placement.

[cjd]

What's the consensus on use of HT receivers room EQ?
I thought most HiFi buffs had contempt for them.
I'm sure that using room EQ is not as good as a rectangular room of the Golden Ratio with speaker placement of Golden Ratio but lets face it this does not suite most HT installations.

I wonder what the lesser of the two evils is,

consumer DSP room EQ
or
unequalized room with non-optimal speaker placement.

I have an Onkyo 705, and can say while the room EQ is mildly useful it sure seems very limited in what it can actually do: to fix a mild peak at one frequency I end up having to notch out around it and hurt overall response because I can't target the problem frequency - it's basically an automated digital slider EQ - at least, so far that's what I'm finding. I've not been able to determine if the automatic mode will target specific frequency and slopes as it should, or if it just uses these generic frequencies anyhow. Probably more than adequate for most folks but ...

I'll eventually get around to room/wall treatments to fix my issue (I believe it's reflections between driver and AT screen, probably also screen frame plays a role).

C

[ThomasW]

hat's the consensus on use of HT receivers room EQ?

At this point the best approach is to do as much as possible with passive room treatments, then fine tune using EQ. Although slowly getting better the DSP build into most receivers isn't adequate to this task. So unfortunately the processing power is just enough to be frustrating. This means standalone DSP is the way to go.

[Dennis H]

I think some people expect too much from EQ, at least on the main speakers. I think of it as a tone control that can adjust for things like BSC (speaker close to the wall or out in free air) but it won't fix a bad room and it won't fix a bad speaker.

[A9X]

Without good measurement gear, and the knowledge of how to use it and interpret results, you are going to find it difficult to build something good, that doesn't fall prey to Beranek's law.

Beranek's Law

It has been remarked that if one selects his own components, builds his own enclosure, and is convinced he has made a wise choice of design, then his own loudspeaker sounds better to him than does anyone else's loudspeaker. In this case, the frequency response of the loudspeaker seems to play only a minor part in forming a person's opinion.

L.L. Beranek, Acoustics (McGraw-Hill, New York, 1954), p.208.

The measurements are objective; you are not (I know I'm not).