Speaker Spec Question: +/-3dB vs other measurement

msg

When I see some speaker specifications, I see a line something like this:

Frequency response
50Hz-22kHz +/-3dB

I've also seen this specced at different dB range, like:

+/-1.5dB

What is the difference in these two specifications in real world evaluation of a speaker's performance? What does it mean?

Is one of these better/more accurate than the other?

pitdogg2

Diff is 1.5 db of course.
Lol

delkal

Seems like +/- 3 dBs is the "standard" spec most speaker companies reference. So you can have a 6dB spread between frequencies and it is still in spec. +/- 1.5 dB's should be flatter and more accurate but whether or not it sounds better is subjective. Most people tolerate (or even prefer) some emphasis or deemphasis at certain frequencies. It can give character to the speakers.

Also make sure you take the frequency range into account. It is easy to fudge the numbers if you cut out the extremes.

Gardenstater

I believe IIRC -3 dB is perceived as half volume to the human ear. I think it's important to compare apples to apples when comparing those freq responses. ie. +/- 2db vs +/- 2db and +/- 3 db vs +/- 3dB. My Polk 7B brochure I got when I bought them says 33 Hz - 20.5 kHz +/- 2 dB. It always makes me laugh when speaker mfg's give a frequency response with no +/- dB's! Like.......ok.....howdya come up with that one hmmmm?

mhardy6647

As in many things in life tighter is better (assuming that frequency response accuracy is important).

The BIG PROBLEM (IMO) is that there's no universal standard cleaved to for measuring, nor for reporting, loudspeaker frequency response.

And here's an extra-skeptical take (from me!?! I know, it's inconceivable...) on the spec. At face value, there are two ways to read "plus or minus 3 dB" as a constraint. It could mean "a variation of not more than 3 dB above, nor 3 dB below, a reference level"... or... it could mean "a variation of up to 6 dB in toto." The 'letter of the law' in the latter case, could be met with a curve that was measured from - 6 dB point to -6 dB point from the midrange (as long as the response stayed at or below the reference/zero level. - 6 dB is actually gettin' to be audibly a fair deviation from 'zero dB'.

To think of this another way, an amplifier rated "50 watts +/- 3 dB" could be in spec at 25 watts (- 3 dB) or 100 watts (+ 3 dB) by the former interpretation, or at 12.5 watts (- 6 dB) or (not and!) 200 watts (+ 6 dB) by the latter.

HISTORICALLY, the writers of loudspeaker specifications were just a little bit below used car salesmen in their perceived forthrightness. On a logarithmic scale, of course

Oh, I'll also mention that, in olden days, a spec called "frequency range" was popular. It was typically taken to mean "-10 dB points"

EDIT: PS The popularity of "-3 dB" as a limit for things in hifi arises from a longstanding belief that a 3 dB difference is about the minimum reliable perceived difference in, e.g., loudness.

Gardenstater

Well I kinda got it a bit wrong. - 10 dB is half as loud (not -3 dB). Perhaps -3 dB was half the power? (< he says trying to weasle out):

https://www.axiomaudio.com/blog/frequencyresponse

"But no speaker is perfect. Consequently, some tones may emerge from a given speaker somewhat softer or louder than others. We measure these variations in the speaker’s output--its frequency response--with the decibel (dB), a unit of relative volume, 1 dB being the smallest change in volume that most of us can detect. A difference of 3 dB is quite noticeable. A difference of 10 dB subjectively sounds like "twice as loud," or "half as loud" if, for instance, the speaker’s output dropped by -10 dB at certain frequencies."

msg

Gardenstater wrote: »

I believe IIRC -3 dB is perceived as half volume to the human ear.

This is kind of what I'm looking for here - a basic explanation to start. This makes sense to me in real world terms.

I tried to do some quick reading before posting this question, and sort of picked up that +/-3dB spec is to say/show that a speaker is "generally flat from x to y" within this +/-3dB range. Apparently this "flatness" is a generally desirable spec, but as @delkal says, some tolerate/prefer some accentuation.

For specific context, if I'm looking at a speaker - a bookshelf, especially - I'm curious to gain a sense for its general bass output performance. I thought the +/- had something to do with its output capability - reasonable performance/clarity at a particular volume level.

Is +/-1.5dB higher volume or lower than +/-3dB?

Does speccing 50Hz at +/-1.5dB exaggerate the speaker's bass output performance vs speccing 50Hz at +/-3dB?

I'm thinking about Audioengine's HD6 in this case.
https://audioengineusa.com/product_tech_specs/hd6-wireless-speakers/

mhardy6647

msg wrote: »

...

Is +/-1.5dB higher volume or lower than +/-3dB?

Does speccing 50Hz at +/-1.5dB exaggerate the speaker's bass output performance vs speccing 50Hz at +/-3dB?

I'm thinking about Audioengine's HD6 in this case.
https://audioengineusa.com/product_tech_specs/hd6-wireless-speakers/

Maybe yes, maybe no.
It depends totally on the conditions under which the measurements were made.
What it should mean is that the output (SPL) at 50 Hz is closer to the midrange reference level by 1.5 dB than were the spec at +/- 3 dB. The corollary of that should be that the bass roll off is a little bit lower for the +/- 1.5 dB spec'd loudspeaker than for the hypothetical +/- 3 dB spec'd loudspeaker. But we're getting into the realm of angels on pinheads to talk about 1.5 dB differences and we're in the range of frequencies where the end-user's acoustic environment will have far more impact (could be 10 or 20 dB, or more, and extremely frequency and position dependent!) than a 1.5 dB difference.

Specs are of pretty limited usefulness when it comes to loudspeakers, I would opine (with my characteristic lack of humility -- or is that humanity?).

msg

mhardy6647 wrote: »

and we're in the range of frequencies where the end-user's acoustic environment will have far more impact (could be 10 or 20 dB, or more, and extremely frequency and position dependent!) than a 1.5 dB difference.

Specs are of pretty limited usefulness when it comes to loudspeakers, I would opine (with my characteristic lack of humility -- or is that humanity?).

Yes, this for sure.

I just didn't know how to take that 1.5dB spec, since I'm used to seeing it at 3dB, and always wondered how this worked, but never managed to absorb the reading. And, of course, as noted, specs to be taken with a grain. Just looking for ballpark sense.

mhardy6647

OK, since I am obsessing about this and since I have nothing better to do and since I don't mind absorbing ridicule on an internet-wide scale -- let's take a look at a sort-of real world example.

Love 'em or hate 'em, thanks in great part to John Atkinson, Stereophile does a pretty good job of providing consistent objective quantitative measurement of the products that they actually test. Whether their measurements are the right ones may be debatable, but they apply their techniques (or tactics, if you prefer) pretty consistently, and they document what they do (and when they make a change to their test regimen), so they give us some kind of reference point to talk about things like speaker measurements.

I have randomly chosen a loudspeaker review from Stereophile in an attempt to illustrate loudspeaker frequency response analysis The 'random' tactic I used -- this was the third one on a 'speaker review' page I clicked on, and it looked like a fairly normal two-way speaker

This is from their review of the Manger P1 loudspeaker published in November 2019.
https://www.stereophile.com/content/manger-p1-loudspeaker

The published specs may be found at https://www.stereophile.com/content/manger-p1-loudspeaker-specifications
These are $15,000 the pair and the manufacturer specifies the "frequency range" (which might mean +/- 10 dB, who knows?!?) as 40Hz–40kHz.

JA's suite of measured data may be found at https://www.stereophile.com/content/manger-p1-loudspeaker-measurements
Remember that a doubling of frequency is equivalent to one octave; i.e., 20 Hz to 20,000 Hz is a ten-octave range of frequencies. The piano keyboard (fundamental notes) covers a range of seven octaves.

Here is a 'synthesized' frequency response for the speaker as measured by JA, including a terse description of the machinations he uses to generate it:

Fig.4 Manger P1, anechoic response on HF-driver axis at 50", averaged across 30° horizontal window and corrected for microphone response, with the complex sum of the nearfield midrange and woofer responses plotted below 300Hz.

Just for fun, I guesstimated a +/- 3dB "envelope" around the data presented in Fig. 4 above. My choice of a "0 dB" reference point is entirely arbitrary!



One thing that should be very obvious from the original Figure 4 is that the speaker's "quas-anechoic" (my words, not JA's) is not flat. This is probably because there's a lot of psychoacoustic research that suggests a 'flat' anechoic response will make a loudspeaker used in the real world sound bright to most listeners. A downward-tilted response curve is felt by some (e.g., Harman ) to be perceived the most pleasant to listen to.

Consequently, I guesstimated the 'line of regression' through the reasonably (log) linear part of JA's frequency response curve:




This is just my pontification rumination.
I'll let you read JA's comments for yourself.

so... how would you spec the frequency response of this loudspeaker?

mhardy6647

oh, I'll add one other meta-analysis.

https://us.v-cdn.net/5021930/uploads/editor/u0/vlexa7rojxxc.png

By my eyeball, the 'flatness' of the frequency response across the range from about 40 Hz to 20,000 in "Figure 4" is about 12 dB -- with a peak (maximum) at ca. 200 Hz and a valley (minimum) at ca. 1800 Hz.

FWIW, I might be tempted to apply the following "spec" to this loudspeaker: "Frequency response: 40 Hz to 20,000 Hz, +/- 6 dB" based (only!) on Fig. 4.

mhardy6647

msg wrote: »

mhardy6647 wrote: »

and we're in the range of frequencies where the end-user's acoustic environment will have far more impact (could be 10 or 20 dB, or more, and extremely frequency and position dependent!) than a 1.5 dB difference.

Specs are of pretty limited usefulness when it comes to loudspeakers, I would opine (with my characteristic lack of humility -- or is that humanity?).

Yes, this for sure.

I just didn't know how to take that 1.5dB spec, since I'm used to seeing it at 3dB, and always wondered how this worked, but never managed to absorb the reading. And, of course, as noted, specs to be taken with a grain. Just looking for ballpark sense.

It should mean the response curve is tighter (closer to the 0 dB reference point) across the range specified.

Gardenstater

@mhardy6647 I think your linear regression line method, with parallel lines above and below determining the "smoothness" of frequency response (as opposed to strict "flatness") is a good one pertaining to listenability of a speaker, provided the slope of the line is taken into account and speakers that have too much of a slope are recognized as deficient. How much constitutes too much I couldn't say.

mhardy6647

Gardenstater wrote: »

@mhardy6647 ... and speakers that have too much of a slope are recognized as deficient. How much constitutes too much I couldn't say.

Oh, there are plenty of folks with strong opinions about that.
https://www.audiosciencereview.com/forum/index.php?threads/jbl-305p-mkii-and-control-1-pro-monitors-review.10811/

Full disclosure: I drank the Kool-Aid and recently joined that forum's throng of "Happy members who like to discuss audio and other topics related to our interest." I blame @clipdat

I am not really one of the folks with a strong opinion about the predictive power of frequency response curves, though -- I like to judge loudspeakers by two things.
How they sound (which I would call "tone" in a ham-fisted appropriation of a rock guitarist's assessment of a loudspeaker!) and how easy or hard they are to drive for the amplifier I want to use 'em with.

That latter criterion, of course, doesn't apply to powered loudspeakers.

mhardy6647

For completeness (and since I just learned that there was such a thing about a week ago ), here as best I can tell is the "Harman Target Response Curve" for loudspeakers (as opposed to headphones). This looks to have its genesis in much earlier work on loudspeaker response characteristics, and I don't pretend to have an encyclopedic view of the topic!



see, e.g.,
http://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html
http://www.audioheritage.org/vbulletin/showthread.php?39134-Harman-Target-Curve
http://techtalk.parts-express.com/forum/tech-talk-forum/1340306-anyone-designing-for-the-harman-target-response-curve
http://www.juloaudio.sk/Umiestnenie_reprosustav/History of Harman Target Curve.pdf

delkal

mhardy6647 wrote: »

Consequently, I guesstimated the 'line of regression' through the reasonably (log) linear part of JA's frequency response curve:

Just looking at the graph those speakers are terrible with a frequency response curve that is a joke. They drift from +4 to -8 dB's, midrange should be lacking, and they struggle to play anything over 11,000 hz. Even old listeners should be able to hear the lack of treble. And the 'line of regression' should not be there either. It should be as flat as possible.

But this is where things get interesting.......these speakers go for $15,000 the pair so I would think someone thinks they are good. Has anyone heard them and (more importantly) liked them?


A few other Random points. The 0dB line is arbitrary. It is usually specd at 1 watt (the same as where the efficiency is calculated) but all you have to do is turn up the volume and the zero changes. But the +/- flatness of the response should stay relative.

To put 3dBs in context. If someone is listening to your stereo and says "Turn it up just a bit" you will probably turn it up 5dB's. Not all people can hear a 3dB change when listening to music (and no one will ever say too much turn it back down).

10 dBs is generally regarded as double the sound volume and it takes 10 times the power to get there.

And watch the axis scale. If you graph using a large spread on the Y axis any speaker can look good. Pay attention to where the +/-5 dB lines are. The frequency response should stay well within those lines. The closer the "better".

halo

Gardenstater wrote: »

I believe IIRC -3 dB is perceived as half volume to the human ear. I think it's important to compare apples to apples when comparing those freq responses. ie. +/- 2db vs +/- 2db and +/- 3 db vs +/- 3dB. My Polk 7B brochure I got when I bought them says 33 Hz - 20.5 kHz +/- 2 dB. It always makes me laugh when speaker mfg's give a frequency response with no +/- dB's! Like.......ok.....howdya come up with that one hmmmm?

Well, the good folks at Polk provide no +/- when they state the frequency response for the Signature Series speakers so you may want to take that up with them

mhardy6647

halo wrote: »

Gardenstater wrote: »

I believe IIRC -3 dB is perceived as half volume to the human ear. I think it's important to compare apples to apples when comparing those freq responses. ie. +/- 2db vs +/- 2db and +/- 3 db vs +/- 3dB. My Polk 7B brochure I got when I bought them says 33 Hz - 20.5 kHz +/- 2 dB. It always makes me laugh when speaker mfg's give a frequency response with no +/- dB's! Like.......ok.....howdya come up with that one hmmmm?

Well, the good folks at Polk provide no +/- when they state the frequency response for the Signature Series speakers so you may want to take that up with them

Assume -10 dB points and one's not probably too far off from reality.

FWIW, here's what Polk published for the 10 and 7 ca. 1977. I'd take the graphs, and the numbers, with a grain (or maybe a large quantity) of NaCl, though.


origpolkmonitorspecs by Mark Hardy, on Flickr
polk_brochure_back_pg by Mark Hardy, on Flickr

I really do need to make a better quality scan of that whole brochure.

Gardenstater

Either Polk made improvements by the time I bought mine in 1979, or their noses got longer: