Do you use a reference when designing a speaker?

[fjhuerta]

This may sound like a strange question...

I just got my Monitor Audio Silver RS8's fixed (they had a blown tweeter). I listened to them side by side with my Monoliths (https://www.htguide.com/forum/showthread.php?t=25871).

My wife commented on how great the RS8s were. When I switched to the Monoliths, we both immediately thought the RS8s were a lot better sounding.

Then again, I measured my design a couple of days ago, and at 15° off-axis, FR is perfectly flat. Power response has a slight downward slope. The RS8's power response is similar to this one (the RS6 power response averaged at 30°):



(From Stereophile).

Anyway, I think that the RS-8's may tend to impress in the short run, and become very fatiguing in the long run. But - this is a big "but" - do colorations can lead to a better sounding speaker? Is flat the way to go?

I realize this is a dumb question - I know a power response should have a slight downward slope - but the Monitor Audios are praised in the press, and their power response is... well... crazy looking (at least to me) I can't imagine what their on-axis response may look like.


Regarding the RS6's measurements:

Below the top octave, the RS6 slopes up gently from 300Hz to 11kHz. When unaccompanied by resonant problems, this sort of response gives the impression of articulation detail and space rather than brightness per se, which ties in with BJR's auditioning comments.

So... is this a correct way of designing a speaker?

Overall, this is impressive measured performance from a relatively inexpensive tower loudspeaker.—John Atkinson

8O

So... I've been designing speakers the wrong way for 2 years?

I'm not being cynical at all... but this "conventional wisdom" isn't exactly what I'm used to reading. And besides, my wife loved the sound of the RS-8s.

Is this a valid way of designing speakers?

[Landroval]

My wife commented on how great the RS8s were. When I switched to the Monoliths, we both immediately thought the RS8s were a lot better sounding.

Can you describe in more detail what was better in the sound of the RS8's?

[fjhuerta]

Can you describe in more detail what was better in the sound of the RS8's?

Sure. The midrange was more prominent, not as laid-back. Bass wasn't as overwhelming. They seemed to be quite a bit more detailed. The RS-8's, for the lack of a better metaphor, were "cleaner" and "lighter" sounding. My design was less detailed and were not as present. I don't know if this makes sense.


This is the simulated on-axis FR of my speakers. It's basically identical to the measurements I took.

[JonMarsh]

Just a few thoughts...

Speakers are often not placed correctly in a room to achieve flat upper to low bass response and balancwe wiht the rest of the spectrum. An uptilt to the SPL curve may subjectively "balance" this, though you wind up wtih a saddle shaped curve. This is also a form of loudness compensation and may make you happier with the playback at lower than "real" SPL levels.

Then there's all the issues around transparency in source components and amplification and cables.

Since I usually use Cardas style setup, and playback at SPL levles closer to life, an uptitiled speaker sounds unnatural and fatiguing after a while, though it may sound more detailed at first.

Monopole speakers will also tend to have a power response that is downsloped because of the difference in fundamental radiation power at different frequencies- this is part of why a well designed dipole is quite attractive to me. Compensating this by raising the on axis response would be OK if you listen mostly in the distant reverberent field (say, in an adjacent room), but not if you listen where early arrival sound dominates. IMO, a design should be worked out that is good in both characteristic. This leads to dipoles, or unconventional monopoles. For example, Avalon speakers are designed with a very small baffle size for each driver, so that the radiation angle is into more than just a 2 pi space. But then the compensate this wtih a slight downslope in on axis and power response. And of course, they are normally setup in a Cardas style room location to optimize balance, minimize room interaction, and extend perceived LF response.



Cardas Speaker/room setup

If you choose NOT to use this kind of setup, then you're inherently picking a colored and incorrect locations for a monpole box speaker, unless it's flush mounted into a wall, nearby walls and ceilings are acoustically treaded to control early reflections, and the speaker crossover is designed for "infinite baffle" (no BSC).

For more thoughts on boundary issues, see Evil twin's post in the reference threads.

Boundary effects on bass response

[jimangie1973]

How about the FR at 30 deg and even 45 deg off axis? Are there any peaks, even small ones in the 1-3 kHz range? I've found that any peaking in the off-axis response correlates to listening fatigue and a sound without that really satisfying something where you know it's right.

In my experience, it's definately an iterative process to get the sound just right. This is true even if you do measurements for impedance and FR in the enclosure before designing the x-over initially. You need a good x-over simulation tool where you can clearly see how the response changes when you change component values.

Another good tool is a 1/3 octive EQ, so you can try different things like BBC dip, or inversing any anomoly you see in measurements, etc. It helps get a target response you can then match the simulation to.

If you're not happy with the sound, keep at it.

[cjd]

In my experience, what happens with speakers designed like that RS6 is that, they grab you first thing. But over an hour or two of listening, your ears start to fatigue significantly. The "meat" of sound is in that dip, while the "detail" of sound is above that (starts ~800Hz, though - if it's got a mild peak 800-3k it'll have "forward" or "detailed midrange" - above that tends to add to "sparkle" and such. It's not about a correct design specifically, but about understanding what type of response triggers what response in a listener.

The RS6 will be more likely to sell when it's on a showroom floor, or even a short in-home demo. It might drive folks with a classical background a little batty.

This is just to add to Jon's comments about listening levels and room setup.

There's never a simple answer.

[TacoD]

Did you measure both speakers (left and right), maybe one of your speakers is wired out of phase.

[mazurek]

I'm going with Jim and Taco, measure your speakers much more comprehensively. Lately, I've been using ARTA, which has a function that averages many measurements. I've been doing measurements over a +/-30 degree window at and a little above and below the tweeter, and shooting for a flat average of all those measurements. Also, I've been enjoying lately harmonic distortion sweeps at multiple levels up to 105 dB/m; I'd say they go hand in hand with impedence tests in finding any box weaknesses, or driver weaknesses. Another thing you can do is measure nearfield, 1/2 meter, 1 meter, 2 meters, and at listening location, which gives you a better idea of room interaction and a rough picture of actual baffle step response.

Another thing, I agree crossovers are a very iterative process. Everyone has slightly different tastes, and may enjoy different listening levels. What works for some may not work for others.

[Jed]

I would measure your Monitor Audio speakers in your room to see what is really happening with the power response and on axis. Also, the speakers you are comparing to them are completely different in regards to planar versus dome, distortion profiles, crossover topology and slope. I'm not sure it is just about the FR because there are so many other variables that influence the sound characteristics even though 2 sets of speakers may seemingly have similar FRs on the measurement axis. Take for example, ribbons, many claim they have a quality that is very appealing in the upper treble that can not be matched by a dome, however both might measure the same. Here the difference in harmonic distortion might influence the outcome, or perhaps it is the way in which the drivers project sound into the listening area/dispersion characteristics.

[fjhuerta]

Thankyou very much for your information. It has been (as always) interesting and very informative.

I've checked so far phase, on axis, 15° and 30° off axis in every design. So far, things work fine.

I do understand how bass reinforcement could be the culprit here - since I'm aiming for a flat power response and I'm not considering the room where the speakers are played in. But, if I do so - they won't work anywhere else... and my wife would kill me if I used my speakers using the Cardas guide.

The Monitor Audios have been playing for a couple of hours now. They aren't tiring at all - I really enjoy them. I'll get used to them, then I'll switch back to my own speakers.

I'll then take some 1/3rd octave measurements in-room for both speakers. This may be enlightening...

Thanks again for your opinons!

[augerpro]

Get a 31 band EQ and start playing with the monoliths. I'll be doing this exact experiment soon. May help to determine your tastes and find if/ at what frequencies extended listening starts to fatigue the ears.

Just last night I was dinking around with the windows media player eq and found some interesting things. 1000Hz bump adds a sort of soaring sound that would sound cool on classical or acoustical music, but it was also sort of honky and unnatural, I didn't like this at all.

2000Hz bump makes vocals clearer. In moderation I didn't find much benefit or penalty, sort of neutral as far as that goes.

4000-6000Hz bump really makes it clear and sparkly and open and all those other magical qualities people babble about.. At first listen this sounds just wonderful. I don't doubt that RS6 sounds great doing quick comparisons, since it has higher levels in this area. My question is: is this fatiguing after awhile? Or is does the quality of tweeter (or suppression of woofer breakup) have more to do with that? I do know that when I first audition some speakers I prefer this brighter, clearer sound. I also know I get tired of it, which is why the last two sets of speaker I bought have been slightly laid back. My question is whether it was the response that I know tires me, or the fact that I've listened to mostly lower end speakers, with their cheap tweeters, poor XO's that don't bury breakups, etc.

[---k---]

I read all of sterophiles' reviews for over a year. At the end of the day, I appreciated the test data but found the commentary worthless. They praise speakers which have an absolutely terrible measurements. If you read JA's comments very carefully, you can tell that he is choosing his words very very carefully.

I think a reference is critical to judging speakers. Had I not heard CJD's MTM, Paradigm Studio 40, Dyn 42, Ascend 340SE all at once, I don't think I would have ever really learned what each really did versus the others.

My Ascends always sound great to me, until I switch back and forth from them to any of CJD's designs. The difference is always immediate, and I've always selected the CJD's designs as clearly superior.

[cobbpa]

I'll then take some 1/3rd octave measurements in-room for both speakers. This may be enlightening...

This should be pretty interesting and very informative! Really getting to see what you prefer & maybe why, understanding what's happening in-room, etc. Looking forward to thos results, please post them if you can :T

[JonW]

Javier,

Interesting question. I've got a pair of commercial speakers that I really like a lot (Totem Forests). They have a frequency response remarkably similar to the graph you showed- a little higher SPL's in the highs and lows, a little lower in the mids. I've been listening to them for about 2 years now and I think they're great. This graph is from Stereophile.



When I built my Modula MT's I was always comparing them to the Totems, going back and forth. I ended up adding in a resistor to the tweeter section of the Modulas to tweak the sound a wee bit, to bring up the highs a little more.

I'm now in the process of designing my own MT speakers. I'm curious to see what type of sound they will have, how it turns out I like them sounding best. I do wonder if the graphs will end up similar to the non-flat Totems. I know that people consider flat to be "correct" and non-flat is "incorrect." But maybe I don't like flat? Not sure. We'll see.

-Jon

[JonP]

Others could elaborate better, but that's very much the "BBC Dip" that is sometimes designed into a response. IIRC, it's a broad, shallow dip, usually more centered around 2kHz, and the overall effect is more relaxed. Kind of a "10th row" rather than a "front row" concert experience...

Someone correct me if I got some of that wrong...

[jimangie1973]

Others could elaborate better, but that's very much the "BBC Dip" that is sometimes designed into a response. IIRC, it's a broad, shallow dip, usually more centered around 2kHz, and the overall effect is more relaxed. Kind of a "10th row" rather than a "front row" concert experience...

Someone correct me if I got some of that wrong...

That's definately the effect. In many designs with a wide dispersion dome tweeter, the BBC dip creates a flatter power response. When the tweeter comes in around 2 kHz, it's off-axis energy is much greater than the mid-woofer. So cutting it's output creats an on-axis dip which fills when measured off-axis.

[JonMarsh]

Nah, you're pretty much on the money. It's an issue because so much stuff is relatively close mic'd. Also there are power response versus axial response issues, and then there's that whole psychoacoustic saddle stuff, and loudness perception. I'm still dithering about whether to put a little BBC dip in the Modula NeoD; I've listened to them with about three different sets of electronics and a lot of different program materail.

For the intended use in HT, I'm not sure it's a wise idea- for music, it might be. The more transparent and natural the source material and the electronics, the less it seems desirable. We'll see....

[kingpin]

So the final design of a crossover is tweaked by ear or by measurements?

Considering that everybody hears something different and has different tastes are the majority of the designs here made to have as flat as a response as possible?

What if you hear the most amazing speaker you have designed and then you measure the response and it doesn't look so good.
Do you go back tweak until it measures flat and then listen again?
Is your brain not fooling you into saying, "ah, if I tweak to a flatter response it will sound better".
Now you are already thinking that tweaking is going to sound better so thats what you automatically hear, wether or not it sounds better or not.

What I am getting at is this.
You design speaker x.
It sounds awesome.
Response is not flat.
Tweak to make it flat or as flat as possible.
Listen to it again.
Tweak more.
Measure again.

What is the your final conclusion. Measurements or what your ears tell you.

Sorry for all the questions.

[fjhuerta]

Javier,

Interesting question. I've got a pair of commercial speakers that I really like a lot (Totem Forests). They have a frequency response remarkably similar to the graph you showed- a little higher SPL's in the highs and lows, a little lower in the mids. I've been listening to them for about 2 years now and I think they're great. This graph is from Stereophile.



When I built my Modula MT's I was always comparing them to the Totems, going back and forth. I ended up adding in a resistor to the tweeter section of the Modulas to tweak the sound a wee bit, to bring up the highs a little more.

I'm now in the process of designing my own MT speakers. I'm curious to see what type of sound they will have, how it turns out I like them sounding best. I do wonder if the graphs will end up similar to the non-flat Totems. I know that people consider flat to be "correct" and non-flat is "incorrect." But maybe I don't like flat? Not sure. We'll see.

-Jon

Jon, great and helpful post. Those are, indeed, my feelings at this point. I suppose flat is nice, but who has a perfect living room? I'd rather have "good sounding" at this point.

I wonder if a resistor, in my case, would tweak FR enough. I need to take those 1/3rd octave measurements first.

[JonW]

So the final design of a crossover is tweaked by ear or by measurements?

Considering that everybody hears something different and has different tastes are the majority of the designs here made to have as flat as a response as possible?

What if you hear the most amazing speaker you have designed and then you measure the response and it doesn't look so good.
Do you go back tweak until it measures flat and then listen again?
Is your brain not fooling you into saying, "ah, if I tweak to a flatter response it will sound better".
Now you are already thinking that tweaking is going to sound better so thats what you automatically hear, wether or not it sounds better or not.

What I am getting at is this.
You design speaker x.
It sounds awesome.
Response is not flat.
Tweak to make it flat or as flat as possible.
Listen to it again.
Tweak more.
Measure again.

What is the your final conclusion. Measurements or what your ears tell you.

Sorry for all the questions.

Mike,

I haven't been through a complete speaker design yet, to be able to say much. But here's one thing to think about: Pick a price for a pair of speakers. Let's say $3,000. Go to stores (or better yet bring them home) and listen to speakers from, say, 5 different companies. They all sound totally different, don't they? Interesting. Lots of design decisions and choices come into play.

-Jon

[JonW]

Jon, great and helpful post. Those are, indeed, my feelings at this point. I suppose flat is nice, but who has a perfect living room? I'd rather have "good sounding" at this point.

I wonder if a resistor, in my case, would tweak FR enough. I need to take those 1/3rd octave measurements first.

Yeah, I'd also like "good sounding." Now that my measurements are done, I can start modeling my crossover. I'm not sure if I should target a flat response, as is typical. Or maybe I'll try one with this BBC dip. Hmmm... Maybe I'll try both and see. I've got the DCX2496 so I can try out multiple configurations.

The "flat" Modula MT's were not quite to my tastes so I changed them very slightly. With regard to resistors, have a read through the Modula MT thread in Missions Accomplished. I explained what I did there to make the sound a little more to my taste. There was also discussion in this thread:

https://www.htguide.com/forum/showthread.php?t=23191

Good luck. :T And, of course, tell us what you find along the way.

-Jon

[cobbpa]

Warning: inexperienced builder rambling based only on what I've read, not what I've done.

Have either of you considered aiming for a response similar (as identical as possible) to the commercial speakers you enjoy? I realize that off-axis & power response and many other factors change the sound, but it seems like that might be a good starting point. Of course, I have never designed a speaker from scratch, so perhaps this is a bad idea. It just seems logical that if you are using (probably) lower-distortion & more capable drivers and crossover components that a similar response plot would produce a similar sound with lower distortion. Both graphs appear to drop nearly 5 db from 100 to 500 hz, then rise back up 5 db at 10 khz. That's very general, but it's the similarity I see. I thought the "typical" BBC dib centered on a frequency an octave or so higher, but maybe I'm recalling a minority of designs. Regardless, I would guess that trying to manufacture that shape across the same frequencies would be worthwhile. Now, I do realize this is all paper racing, conceptual and a plethora of variables determine sound, so it may not be likely that the individual drivers used end up sounding as pleasing with that same dip. I realize this is what Javier was aiming for and asking about, I'm just throwing out my (note: unexperienced) opinion that perhaps copying the on-axis response would seem to be a good place to start in an attempt to find out what's so appealing. I think it would be interesting to hear how dissimilar the two speakers end up sounding.

In summary: an inexperienced reader votes to try and copy that curve :T

On a side note, Jon, have you measured your Modulas with the resistor tweak? I remember preferring the Totems, even with your modification. I'm curious as to how similar the responses are now.

[audioslave]

bass hump....

From what I have read in the Stereophile reviews, JA usually comments that the 6db rise in the bass is due to his measurement technique. I see that he didn't make that comment in the RS6 review, however he noted that the bass alignment was very complex (i.e. twin ports with different tuning for the midrange and woofer).

Other things to note that differ from many other sets of measurements, is the smooth off axis response in the horizontal and a below axis suckout at near the x-over freq. A lot of other subjectively impressive speakers seem to have on axis suckouts that are filled in off axis horizontally and x-over suckouts above axis vertically.

All in all it is very difficult to find much correlation in the measurements to the subjective reviews.

[Jim Holtz]

My take on this subject, FWIW...

I've been following this thread reading the various opinions about why different speakers sound good to someone when they don't measure flat. Here's how I see it.

IMHO, what you're really talking about is the "black art" of speaker building, voicing. The reason all of those different speakers sound different is because they were all voiced differently to suit the designers tastes. Personally, I want a speaker to sound as accurately as possible. The closer it takes me to how I perceive the live performance, the happier I am. Flat is not always accurate. The crossover topologies and driver differences will influence the sound substantially and often require response shaping to create that accurate sound or the sound the designer "likes" which often has nothing to do with accuracy of music reproduction and is only one of the reasons there are so many different sounding speakers.

I chased my tail for many years as an audiophile trying to find the speakers that had the "slam" I liked and the "presence" I thought was enjoyable, yada, yada, yada. However, I also found that my tastes constantly changed from day to day etc. so I was never really happy with what I had. I finally realized that I had no reference to use unless I compared the sound I was hearing to how I perceived the live performance to sound. Notice, I use the perceive word a lot when talking about a live performance. I wasn't there during the performance and can only guess what it should have been like, but I have an image in my mind of how I think it should have sounded.

Anyway, I'll quite rambling and get off my soapbox by saying, voice your speakers the way you like them but realize that unless you're shooting for accuracy, you may be the only one that does like them. That's what DIY is all about. Building speakers to suit you.

My $.02 worth...

Jim

[---k---]

I've heard people claim that high end head phones are very accurate. Have you ever done a comparison to of your speakers to headphones.

I just have a problem with the whole notion that non-flat is better. To me, that is an acknowledgment that the recording has flaws that must be corrected by the speakers.

However, flat on axis is never flat off axis. I believe that I've read the BBC dip results in better off axis performance. So, I guess it is all trade offs and what your definition of flat is.

[JonMarsh]

I've heard people claim that high end head phones are very accurate. Have you ever done a comparison to of your speakers to headphones.

I just have a problem with the whole notion that non-flat is better. To me, that is an acknowledgment that the recording has flaws that must be corrected by the speakers.

However, flat on axis is never flat off axis. I believe that I've read the BBC dip results in better off axis performance. So, I guess it is all trade offs and what your definition of flat is.

Actaully, a well designed speaker can be flat both on and off axis, but it takes a fair amount of effort and is another design restriction to consider. In practice, few commercial speakers work very hard to do this, but the best do. Well designed dioples tend to have an advantage there, if the figure 8 pattern can be maintained consistently with frequency, but then the upper end is a problem (like the Orion, which with the tweeter flare doesn't maintain the same pattern control off axis, which is part of why it has, guess what, a BBC dip. I'm pretty convinced that the best treble for a dipole is a waveguide with good pattern control over the whole operating range, and a dispersion that reasonably approximates what the open back midrange and woofer are doing. Something like this....



Of course, equalization is required to shape the response to "level", but note the tightly controlled spacing of the response at different angles- there's a consistent fall off at 30 and 45 in level, but that's the point- consistent with regards to frequency. Direct radiator tweeters don't do that, and neither do most horns or so called waveguides like the MCM used in JKrutke's one design.

[JonW]

Have either of you considered aiming for a response similar (as identical as possible) to the commercial speakers you enjoy?

I can try it with my current project. Maybe a xover with and one without a BBC dip. Plug them into the DCX2496 and see what happens.

On a side note, Jon, have you measured your Modulas with the resistor tweak? I remember preferring the Totems, even with your modification. I'm curious as to how similar the responses are now.

No, I haven't had a chance to do the proper side-by-side measurements. But after my recent difficulties with measurements for my current MT project, I won't be doing that again unless I have a very good reason. (OK, I'll be doing it again when I build my final cabinets.)

[JonW]

Those plots that Jon shows are pretty impressive in terms of on versus off axis. Nice.

I just have a problem with the whole notion that non-flat is better. To me, that is an acknowledgment that the recording has flaws that must be corrected by the speakers.

It's not necessarily fixing flaws of the recording. It could be fixing flaws of your ears. They ain't linear. :P

[Dennis H]

SL's reference.....

http://www.linkwitzlab.com/reference_earphones.htm

Short version, he goes to concerts and makes his own recordings using 2 mics on each side of his glasses. Then he compares the recording played back through the speakers to what he remembers of the live sound.



He likes using earbuds/headphones as a reference but most need EQ to sound real using his live/recorded comparison. One he found that doesn't need EQ is the Shure E2/E2C for less than $100. Caveat, earbuds and headphones depend somewhat on the shape of your ear canal so what sounds good to him may not sound good to you. Still, the Shures look like a good buy as a reality check for how well your speakers are doing.

[cobbpa]

Wow, lots of interesting stuff from those far more experienced than I...makes me feel like I'm the same as the annoyings kids in classes who raise their hands far too much and don't contribute anything worthwhile. I'll keep quiet and take my notes now ops:

[fjhuerta]

I think I just found the holy grail!

IIRC, Jon Marsh posted this link some time ago. I just re-read it. It goes into detail into everything I'd ever want to know about on-axis, off axis and power response...

Loudspeakers&RoomsPt2.pdf

I just found a lot more of respect for Harman.

[augerpro]

page cannot be found for me?

[fjhuerta]

page cannot be found for me?

That happened to me the first couple of times, too... and I can't upload it here. The best way to get it, it seems, is visiting www.harman.com, going into the site map, looking for "White Papers" and searching "Loudspeakers".

[TacoD]

Those white papers are really worth reading, those listening test are very informative.

[JonW]

I think I just found the holy grail!

IIRC, Jon Marsh posted this link some time ago. I just re-read it. It goes into detail into everything I'd ever want to know about on-axis, off axis and power response...

Loudspeakers&RoomsPt2.pdf

I just found a lot more of respect for Harman.

I just gave that a quick read through. Their experience indicates that people like to hear flat frequency responses. That's flat when all resonances, reflections, etc. are considered. And their "not flat" comparisons had wavy/bumpy plots. No mention of BBC dips or gradual changes in the frequency profiles, such as raised lows and high/supressed mids. Interesting.