What the 'professionals' think makes a great speaker ~~~

[<HAL>]

Some images of measurements made on loudspeakers that greatly outsell Harbeth and are found in some pro studios. Therefore they must be right! They tell the sound engineer mixing the CD you are willing to buy everything he/she needs to know .....

Points to note = = generally elevated presence band to punch image forward like picture 3 here. A big peak at 4kHz, where the ear is maximally sensitive. Extremely fatiguing to listen to. Apparently endorsed by professionals.

[Don Leman]

If the pro studio's use monitor speakers that have peaks in the frequency response, how will that affect the recordings that come out of their studios? One might assume the recordings would have a dip in response where the peaks were when the project was mastered.

[A.S.]

This is an interesting one. I've looked a bit further into this and for your interest attach a picture of three "pro" monitors (A, B, C) and one "domestic" monitor (D) measured under controlled conditions. D was measured using a different audio test set, but I've made sure the vertical/horizontal scales are directly comparable with the others. (I could possibly have exported D and imported into the later A-B-C measuring system but that needs some time).

What do you make of these? These were measured three years ago and as far as I know, A and B are current models. A is what I'll call a 'Euromonitor'. B is the manifestation of a particular approach to speaker system design. C .... any thoughts? And D? Which one would be expected to give the 'most natural' sound? And if you were a professional sound engineer, based on these curves alone, which one would you have most confidence in as telling you what is really going on in your mix? And which would be likely to give you most or least listening fatigue?

As you can see there are, even now 2011 some eighty or so years after recognisable loudspeakers appeared, very big measurable differences between speakers. Let alone sonic differences.

No competitive brand names will be revealed.

NOTE: Ignore the exact dB markings on the left scale. dBs can be usefully relative (as well as absolute) measurements, like percentages and speaker designers tend not to need to know what a sensitivity figure is, for example, 85.5dB at some frequency. We need to know by how much it varies across the audio band. So, although the left dB scale axis is exactly comparable, you cannot deduce anything from reading the absolute numbers: we've just laid these four speakers out on the page neatly, not in sensitivity ranking.

>

[Labarum]

D has the flattest response by a big margin.

Is the notch at about 180Hz in B down to poor crossover design and the 80 Hz hump in C a booming port?

A is flat enough in the mid bass till the driver begins to roll off but the treble response looks rocky.

A and B have a rising response over 10kHz - could they scream a bit?

Please tell us which of these four designs is active and which passive?

[HUG-1]

Please tell us which of these four designs is active and which passive?

All are passive. C is the largest, probably B the smallest?

Interesting observation ..... 'A's treble response looks rocky'. With some basic tools and no more than five minutes free time could you investigate that issue on A? Before you reach for the tool kit what would your method of attack be to expose the peakiness?

BTW; A, B, D are two way models. C is a three way.

[Labarum]

All are passive.

Then we are not talking about mainstream products of choice used by Audio Professionals, where a passive speaker is today the exception.

In what price bracket is each of the speakers referenced?

{Moderator's comment: can't say for sure. Two examples are heavily marketed to pro users/very serious audiophiles. Don't think the performance would be any better/worse even if an amp was built into/onto the cabinet.}

[EricW]

D would clearly be the most accurate (I think); but it looks as though C would be unfatiguing to listen to because of the rolled-off treble.

I think Don's point above is an interesting one. I would think that a mix made on either A or B, if it sounds right on those speakers, would inevitably sound a bit "dull" on a truly flat loudspeaker, e.g. D.

[Don Leman]

Although it was mentioned only C is a three way, D looks suspiciously like the Monitor 40.1 so that would be my choice.

Of the others I would say that peaks in response are more disturbing that dips. B would be my choice as the rising response above 10k is largely outside the fundamentals of most music so might be noticible as added sparkle. It does have a few dips in in the midrange but I don't think they would bother me as much as the peaks in the other graphs.

{Moderators comment: no! D is not M40! Far cheaper\smaller.}

[DSRANCE]

My own tuppence worth...

Most small "pro" monitors (and I suspect, many larger ones too ), have the presence and treble region peaked up a little so that the engineer can hear distortions and so on in the mixes "under a magnifying glass." These speakers are NOT for comfortable music listening enjoyment and neither are they designed for "total" and "clinical" full-range neutrality - neither are Harbeths IMHO, but I digress and shall need to address this comment later ;) By the way, many European studios and mastering suites use B&W 800 series speakers, which have more of a presence-recessed balance if anything.

I also feel that engineers can get atuned to the character of the monitors they're using, so speakers which measure the way shown above can be got used to and heard "through.".

I was very recently chatting by email to a chap who does video production for a living and is a keen music and vintage audio enthusiast. His preferred little monitor is a UK/far eastern active model with built-in preamp and DAC, which measures very well and apparently sounds extremely good. However, he has extensive experience of the smaller modern monitors that are freely available on the pro market at very attractive prices (and also a few "designer" ones as well for a good few thousand quid), and finds that although they're great studio tools, he couldn't relax with them at home and some, active or not, aren't very clear in sound either (there's a rather popular little active "Euro-box" which turns up time after time in pictures, apparently with this characteristic).

Back to my possibly controvertial comment above regarding Harbeths, and hopefully to bring this post full circle. These days, after many years of listening to all kinds of speakers, good and bad, all I want to do is to kick back and lose myself in the music I choose to play at home. I DON'T want the speaker to throw it at me "warts and all" and neither do I want a pretty false sense of "space" and "depth" which some carefully tailored domestic speakers do (vinyl has often done this all by itself when compared to the master recording in my experience). I want a realistic balance with great and natural reproduction of acoustic instruments from a speaker perfectly able to "rock it" at reasonable levels on occasion, but with great subtlety at lower levels. I find the various "BBC Legacy" speaker types far more able to do this and, of course, Harbeth do seem to me to be leading the field by a country mile at present (the other UK maker with roots in this thinking does well in the UK dealer-chains with their less expensive models which nod to their past, but don't really follow it IMO and their classic series isn't promoted here at all).

I appreciate Harbeth make "pro" based models in their range too, but maybe a "proper" speaker isn't what's wanted in a busy studio where everything has to be done quickly and LOUD!!!!!. I love the 40.1 deeply and had the pleasure of some Monitor 30's a few weeks ago. For me though, the return to the SHL5's brought a sigh of relief, although I fully appreciate the "extra warmth" of the latter may not be ideal in some rooms, but the sound was so much more "comfortable," and wanting me to listen to more and more tracks (I was only there for a day, so much different music to listen to). Having said this, I shortly have some good working LS5/9's coming my way for nowt that spent their lives in a video-editing suite and I eagerly look forward to hearing what the "different" bass alignment will do in my little sitting-room.

I suppose that what the jottings above mean, is that I'd like the clarity of a good pro-monitor, but presented in a gentler, saner and more music and ear-friendly way. This is why Alan spends hours and HOURS listening and fine tuning his designs for long term listening enjoyment, rather than the clinical, magnified analysis of the signal fed them. You CAN heve the best of both worlds

I hope you all take the sentiments above with the sincerity I wish to convey and that I haven't gone off on a tangent with them. Words can be so difficult to describe feelings I find...

[cornelius]

OK - had a quick look, I'll give it a try:

A: the dip between 1-4k will make the monitors sound smooth - home theatre boom and sizzle sound - bass and treble is accentuated - but where are the mids ("compressed" sound?). The peak at 4k will give the illusion of detail and be effective as our ears will still hear that high frequency well.

B: the dip around 200Hz - hmm... "cleans" up the bass so they're easy to place and make one think that they're getting "tight, defined" bass that's "accurate"? Can't understand a dip, most monitors are small and bass is the first thing to go - If anything I'd expect a bit of a bump up, so they don't sound too thin.

C: at first I thought C was a combo of A and B - big dip around 200Hz but instead of a sharp rise at 4kHz, there's a dip - hi end not exaggerated here. 2 interesting things between 200 and 500Hz - not sure what that would sound like. Looks like these speakers go deeper than A and B, these will feel bass heavy (dip at 200Hz and the rise between 50 and 100Hz).

D: obviously the flattest response - does this mean they're the best?

{Mod comment: Edited all to Hz, by convention.}

[A.S.]

....(snip) C: at first I thought C was a combo of A and B - big dip around 200Hz but instead of a sharp rise at 4kHz, there's a dip - hi end not exaggerated here. 2 interesting things between 200 and 500Hz - not sure what that would sound like. Looks like these speakers go deeper than A and B, these will feel bass heavy (dip at 200Hz and the rise between 50 and 100Hz).

D: obviously the flattest response - does this mean they're the best?

A valiant effort to attribute various subjective sonic effects to excesses or shortages of energy in various audio bands. It theory, you're right: we know since 1938 (table attached again from our original thread) that, as you say, even before WW2, a common subjective language was being applied to loudspeaker evaluation. But that's only part of the story. The real devil is what we call coloration and this wretched coloration matter seems to have a dimension to it which is simultaneously intangible and highly significant. That's just the sort of problems engineers shy away from if life and limb are not under threat. I'd like to treat coloration as an entirely separate post because it defines how we judge speakers once we've explained away - as you've started to do - the lumps and bumps in the frequency response. So back to the solely measurable, objective evaluation for now.

In post #5 it was asked ...

With some basic tools and no more than five minutes free time could you investigate [the peakiness in the treble] that issue on A? Before you reach for the tool kit what would your method of attack be to expose the peakiness?

That was a good question because without much effort we can dig around in the speaker system and at least identify how that peak comes about. Since the peak (or indeed any other features of the frequency response) were definitely picked-up by the measuring microphone in the chamber (see picture of Derek accurately positioning the microphone with a large T-square) and they are 100% repeatable, we must be able to determine their real physical source. I'd really like to let you think about this. Remember, just with a few tools and a few minutes, what can we glean about the peak. Or indeed any other curiosity of the measured response. And outside the chamber, two audio measurement systems being compared side by side, fed from the same precision microphone (picture).

BTW, let's assume that we just borrowed these speakers and destroying them in the interests of scientific curiosity isn't an option. And certainly not the 'deep probe' investigation you see in the other picture - although that may ultimately be necessary!

>

[DSRANCE]

That narrow-band peak hints at a metal cone taking off, as most do around 4 to 6kHz as I remember. To tame this, very steep filters set too low for many tweeters (back then) would be needed and a cone peak would be replaced with a rather stressed tweeter harshness..

[A.S.]

That narrow-band peak hints at a metal cone taking off, as most do around 4 to 6kHz as I remember...

Ah ha! Astute answer, but the wrong one! You fell into my trap David!

Can't move ahead on this one until someone tackles this one:

With some basic tools and no more than five minutes free time could you investigate [the peakiness in the treble] that issue on A? Before you reach for the tool kit what would your method of attack be to expose the peakiness?

If you work this through for yourself the result will be very interesting I'm sure and directly relates to your otherwise normally apt comment, but wrong in this case.

Clue. You need a screwdriver if you are unlucky.

[DSRANCE]

Ah ha! Astute answer, but the wrong one! You fell into my trap David!

Just remembering a long-ago chat and comparison of two crossover points on prototype 7's Alan

Screwdriver? Can't be loose drivers/crossover bits can it????? I know that some cheaper tweeters could spit and fizzle more if the fixing screws (not just the mounting ones) weren't reasonably and firmly tight.

[A.S.]

... Screwdriver? Can't be loose drivers/crossover bits can it????? I know that some cheaper tweeters could spit and fizzle more...

Ho ho! You've done it again! You're still in the trap! Tweeter? Bit of lateral thinking please!

[A.S.]

I admit to being disappointed. Let's step back ...

Back in post #3, we showed the frequency responses on axis of four so called monitor loudspeakers. A, B and C are or have been heavily marketed into the professional studio/broadcast markets. D is sold as a 'domestic monitor' and as someone noted, D has the flattest frequency response but is not marketed to the pro market. That's strange don't you think?

Then attention was turned to the peak in the frequency response of speaker A. Contributors suggested that as this was in the 4kHz region (i.e. above the normal crossover frequency for a two way speaker) that the source of the peak was the tweeter itself. I anticipated that would be the observation and asked how that could be verified with simple tools and five minutes. So here is the answer which is obvious (to a speaker designer) ...

1) The peak is real. It is not a measurement artefact. It appears not only on axis, but laterally (and probably above, below too) and will exist today as it did in 2008. In fact it may even have grown in magnitude over time.
2) Where can this peak come from? It's not a subtle notch or bump: it's thin and prominent. It can only be from ...
- i) the tweeter
- ii) the cabinet or grille
- iii) the woofer (?)
- iv) the crossover

or some combination of these perhaps. And it is at a very critical frequency and is most likely to be highly audible because of its magnitude and frequency. (More on that later when we've worked this through).

And the tools we need are - initially our fingers. If the speaker is bi-wired, our luck is in: we can just remove both of the links plug the red/black signal wires into the lower terminals, and without repositioning the speaker or disturbing it in any way, we run a frequency response measurement driving only the woofer via its internal crossover, then save the data. Then pull out the wires and plug then into the upper, tweeter-only terminals. Then we can measure the tweeter via its crossover and save the data. By now we have three on axis measurements with the speaker and microphone is exactly the same positions in space:

- System response, the complete as-sold speaker
- Response of woofer via its crossover alone
- Response of tweeter via its crossover alone

And if we are not so lucky and the speaker isn't biwired, what do we do then?

[EricW]

And if we are not so lucky and the speaker isn't biwired, what do we do then?

I'm guessing that by comparing on axis and off axis measurements at the peak frequency of the resonance, and comparing the measurements with those of adjacent frequency bands, would help isolate the source of the resonance. That is, you'd likely have one type of roll-off if it were the tweeter, another if it were the woofer, and yet another if it were the cabinet. I'm not sure you could isolate a crossover problem this way, though.

[A.S.]

Perhaps. But first we need to capture some acoustic measurements.

So, no biwire connections then. What to do? We unscrew the bass unit, pull off one wire, refit woofer and measure the tweeter + crossover freq. response. Then replace wire to woofer, pull off one tweeter wire, refit tweeter. measure bass unit + xover alone.

Now and only now do we have some frequency plots to even begin to analyse where this prominent peak is coming from.

Phew.

[EricW]

Phew.

I feel a bit dense.

On the other hand, perhaps we've all now been so thoroughly conditioned not to open our Harbeths, we unthinkingly apply the same to other speakers as well!

[A.S.]

Right, let's move onward. Finally - and I'm sure of this - if you're interested in the finer points of why speaker may measure and sound different the ultimate point is likely to be an eye-opener! You won't read this sort of stuff anywhere else!

Let's assume that by now we have some frequency response measurements of the speaker complete, bass unit (via crossover) and tweeter (via crossover). I said we had five minutes and some basic tools. We've fully used the five minutes (and maybe a little more) and I'm hopeful that the measurements we made in haste just by unlinking the biwire links or pulling off the drive unit wire (assuming they are not soldered on) will tell enough of a story. But if we were really doing a serious 'forensic' investigation of a bought speaker we'd have to bypass the crossover completely, and measure the frequency response of the raw drive units when driven directly from the amplifier, still in the enclosure. If the speaker is ported, that will be a big help because we can run the amp cables through the port hole and attach them directly to the woofer and tweeter in turn, having disconnected the normal wires from the crossover that feed them. The wires may prevent the grille from being fitted, which will means that we won't be able to directly compare measurements because all grilles have some influence on the frequency response to a greater or less degree. But it will give us the big picture.

Alternatively, if we actually own the speakers under examination and are willing to permanently modify them, we can drill some holes on the back of the cabinet and run cables from the drive units out through the back to a connector arrangement, and also from the crossover output through the back. Then we have complete flexibility. With no effort and without disturbing the position of the speaker at all, we can just route the amp drive signal to the drivers directly*, through the crossover to the drivers individually or as working pairs. That's really the best way, but it will prevent us from selling-on the speaker later due to the holes in the back panel. We could of course, just remove the drive units from the enclosure and measure them in space, but that will tell us something different and we're trying to dig into this purchased speaker as it is not redesign it. So, whether or not we have the raw drive unit (driven directly) frequency responses or not depends upon how curious we are about deconstructing the system.

We know from this TeckTalk that the crossover has several functions, and just dividing the audio spectrum between the bass/mid driver and the tweeter is only one. Level adjustment and baffle-step compensation are as vital. But what an analogue (conventional, passive) crossover can't easily do - and perhaps shouldn't even be asked to attempt - is to try to cancel-out sharp peaks that may occur in a drive unit. Those sharp peaks are indications that for mechanical reasons that drive unit finds it easier - hence is more efficient at - producing certain frequencies than others. We expect a natural rise in sensitivity (sound pressure output, dBs with frequency) across the working band of a drive unit (then a fall-off) and a crossover/equaliser can restore the sound output to flat over a wide audio range, but cancelling sharp, narrow, tall (high efficiency) peaks is much more difficult. And my well leave a tell-tale sonic signature that sounds as bad or worse than the unattended peak.

So then, what is the drive unit designer's nightmare? If he's designing a bass/midrange unit, he's expecting the device to operate well from, say, 20Hz to 5kHz or more. 20Hz to 200Hz is one octave. 200Hz to 400Hz another. 400Hz to 800Hz, another. 800Hz to 1600Hz another. 1600-3200Hz another and finally 3.2kHz to 6.4kHz another. So that drive unit is being asked to perform perfectly over six octaves. That's a really, really tough technical challenge. That's like designing a car engine to perform in a perfectly predictable linear way from idling to 10,000 rpm without rattles, vibrations, lag or buzzes.

At the bottom end of the audio range the drive unit just has to pump air and virtually any old driver with any old cone material/surround will do. But as we enter the hundreds of Hz and even more critically, up near the top end of the bass/mid driver's range very very strange and complex interrelationships between the electro-mechanical parts become noticeable in the frequency response. The seemingly humble, uninteresting surround (surely that's just a rubber ring that hold the cone to the chassis?) and the cone itself - even the humble, boring dust cap - starts to behave not as one glued-together whole system literally playing the same tune, but each starts to generate sound independently. So the dust cap, for example, can 'sing' at some frequencies even though its made of soft cloth and securely glued to the cone. And taming these various independently vibrating parts is the same problem as trying to trace a rattle in your car which is manifest only at a certain speed - where you think it's coming from and where it's actually coming from are often different places.

OK so far?