It is a way to understand quickly a speaker properties, how it will
sound. It is defined in
this standard
from the US CTA. Here is an example:
The spinorama gives you a few graphs:
On Axis: this the frequency response of the speaker on axis.
You expect it to be as flat as possible above 100Hz.
Listening Window: this an average of various measurements
around on axis. The idea is that it is closer to real life since
your head is always moving a bit in practice and you are not staying
frozen in front on the speakers. It is expected to be close to the
previous one, slighlty going down.
Early reflections: expected to be a smooth, slowy going down
line.
Sound power: expected to be a smooth, slowy going down line
(slope is greater than Early Reflections for classical speakers).
ERDI and SPDI: are the difference between the Early
Reflection curve and the Listening Window (resp. Sound Power and
Listening Window). This should be as straight as possible above 100
Hz. If the DI curves are linear, you will be able to correct the
frequency aberations of your speaker easily. It is important to
understand that a bad directivy cannot be corrected with an IIR eq.
The slope of the line gives a good idea of the speaker properties.
Note that you will not get a flat line except for speakers designed
for this (and they are usually large): they are usually labelled as
constant directivity speakers.
The speaker above is very good. You can compare with another
one:
which is not flat at all. Remember that ±3dB means that the
volume doubles/halves at the corresponding frequency. If you look at
the on axis measurement below, you see that it going up, the speaker
will likely be bright and emphasis high frequency.
Please tell me more:
If you like to read:
You can go to the source (the standard document) and starting page 56, you have a good and compact explanation of
each curve.
The measurements are collected on the internet. Some are of very high
quality and done by independant reviewers, some are of medium or low
quality or provided by vendors.
You can filter the results by clicking the wheel on the website near
the search bar.
How precise are the measurement data?
High quality data: provided by a Klippel NFS or measured in a
large anechoic chamber. Data precision is around
1%. Smoothing post measurement can be done by the Klippel
software so you cannot necessary compare two Klippel
measurements.
Medium quality data: medium size anechoc chamber, data
precision decreases <500Hz. By how much is unknown.
Low quality data:
measured without an anechoic chamber but using windowing for
high frequency and plane ground method for low frequency.
Precision can be as good as a Klippel measurements.
provided by vendors: usually smoothed data (less measurement
points, or data is smoothed, or quality is just
unknown).
Impact of smoothed or poor quality measurements.
Smoothing
generates on average
a higher core higher (+0.7) with respect to a
score computed from the raw
data. Note that the 0.7 can change depending on the amount of
smoothing and other factors. Personally I never compare a score
computed with high quality data with something else.
How precise are the computed data?
Generally precision depends on the quality of the input data. If you
want to compare 2 measurements, you should select them from the same
category.
Score computations
The score is not significant ±0.5. It means that 2 speakers
which have a difference in score less than 1 are in the same
broad bucket. No point to look at the decimals.
Anechoic EQ computations
EQ is for one measurement. Since there are variations speaker to
speaker, you should not look at very sharp or very precise EQ.
How much variation between speakers? This is very brand/model
dependant. Some manufactures provide tolerance like Neumann or
Genelec.
IIR eq (aka PEQs) are not working if the phase is varying a lot
over some frequency range. Group delay (phase derivative) must
be relatively flat. That may explain why EQing around the
crossover frequency does not necessary give excellent results.
On this other side, we are less sensitive to phase anomaly than
frequency anomaly.
Can I use this data to decide which speaker to buy
See dedicated section below.
Should I use room EQ?
Absolutly! Dirac, REW, Audyssey etc will provide audible improvement to
the sound, in general making it flatter (so more tonally correct).
Almost all AVR have one built-in. All computers and most phones can do
it with a built-in or free application.
Focus or broad radiation pattern?
Speakers do not radiate uniformly. Some speakers are designed to be very
focus: they minimise the reflections but the sweet spot is very tiny.
Some speakers are designed to diffuse sound, research proved it is
something that people usually like. If you are near field, for example
on a desk, low directivity works well. The farther from the speaker you
are, the more good directivity becomes important. If the room is
reflective, this is especially important. If you live in a modern flat
with hard walls and floor to ceiling windows, looking for speakers with
a narrow directivity can be a very good way to solve room issues.
How can I help?
This website is generated from
this code
and data from various websites (see below). You are very welcome to
contribute.
You can provide feedback (esp. on bugs, UX, data errors) or add more
datas at
github
or in this
thread
at ASR.
How to select your speaker?
That is a polemic section, I know, I know.
Criteria
I believe there are a few criteria in whatever order you want:
Price:
Price is not linked to quality (sadly) and is a factor where you
are the only judge of the worthiness.
Design:
Design is to each is own. Some like monkey coffins, some like
studio monitors, some like big horns, it is up to you
Build quality:
Some brands built to high standars some don't. Do some due
diligence.
Support quality:
Who can or will repair your speaker if you have an issue? How long
do they store parts? How long is the guarantee? What is covered.
It is worth a close look if you buy once in a lifetime expensive
speakers.
Tonality:
Here we have more concensus, a lot of people like the same thing.
To be HIFI, you want to have a speaker with a high tonality score.
This score takes into account flatness, directivity and how much
bass you will get.
SPL aka Sound Power level:
How loud do you want your speaker to be? It depends a lot of your
room and at which level you want to listen.
Some examples:
Bedroom 3m x 4m, listening distance <2m, 73dB reference
level, +20 dB for peak: 93dB at 1 meter. Around 90dB per
speaker.
Living room 6m x 8m, listening distance <3m, 73dB reference
level, +20 dB for peak: 93dB at 1 meter. Around 96dB per
speaker.
Large room or studio 7m x 12m, listening distance <5m, 73dB
reference level, +20 dB for peak: 93dB at 1 meter. Around 116dB
per speaker.
Bass extension:
Small speakers dont have much bass. You need a subwoofer or a
large speaker if you want both deep bass and high volume. New
speakers with DSP can produce a lot of bass while being small but
not at high volume. If you listen to music only and do not like
electronic music, you dont need subwoofers with large speakers.
Some examples:
Bedroom 3m x 4m, one 10 inch subwoofer.
Living room 6m x 8m, one 15 inch subwoofer or two 12 inch
subwoofers
Large room or studio 7m x 12m, two 15 inch subwoofers, may be
more
You can choose floorstanders or bookshelves with subwoofers.
Integration is not easy but very doable especially if you use an
AVR. Most people like bass, your neighbourgh may not like
your bass.
How good is your room?
Is the room dedicated to music?
Is the room symmetrical?
Does the room have some absorption?
Did you add panels to control reflections?
Did you have multiple subwoofers to control bass linearity?
Did you measured your speakers in your room?
If you did not answer yes each times, then you do not need the best
speakers in the world and obsessing over scores, SPL etc will not make
it the best room in the world. At the same time, I understand very well
the attraction of having a great pair of speakers. Speakers with a high
tonality score will be easier to EQ and will adapt well to your room.
Define the SPL you want (at 1% distorsion) and now you have reasonable
choices.
How to interpret the numbers on the landing page?
Each speaker measurement shows up in a box similar to:
Price: should be self explanatory; it is in USD per
speaker not for a pair. The price was the first hit on Google
when I looked it up. It may be different in your country and/or
later in time.
Tonality: this is a value between -10 and 10. It is defined in
the CEA2034
standard and is computed from the spinorama data.
Higher is better
It make sense for tower, bookshelves or center but
not for surround, in-wall or column speakers; if you
see *** instead of a number, it is to remind you that
the score is not valid for some shapes of
speakers. If you go to the page of a speaker you
will still still see the computed value. It needs to
be taken with care. It may "work" for some speakers
but not for others. The predicted in-room response
assumes a rectangular room with standard reflection
and a dipole speaker. All bets are off for a stadium
or an omnidirectional speaker.
Note that a difference between two scores is
only significant if the difference is greater than
0.6.
Be also mindfull than smoother or less precise data
can yield a higher score than an Klippel generated set
of data for example. You can easily remove 0.5 to 1.5
to the score for highly smoothed data.
The score does not tell about maximum output or
about distorsion. You can get the same score for a
portable bluetooth speaker and a large tower.
If you know how to read code, you can find my
implementation in
python here. There
is also a version in Cython which is faster if needed
be in the same repository.
Bass extension: this is low frequency extension
(LFX in short) from a research paper from Olive. It is computed
as here. Lower
is better. Note that it is slighlty different from the frequency of
the -3dB point. The rational for using it is: The low frequency extension (LFX)
LFX = log10(xSP-6dB.re:y _ LW(300Hz-10kHz) (7)
where LFX is the log10 of the first frequency x_SP below 300 Hz
in the sound power curve, that is -6 dB relative to the mean level y_LW
measured in listening window (LW) between 300 Hz-10 kHz.
LFX is log-transformed to produce a linear relationship between the
variable LFX and preference rating. The sound power curve (SP) is used
for the calculation because it better defines the true bass output of
the loudspeaker, particularly speakers that have rear-firing ports..
Flatness: this is the variation in dB around the average SPL of
the on axis measurement between 300Hz and 5kHz. Lower is better.
Source of data and citations
AudioScienceReviewaka ASR: it is a fantastic source of speakers data thanks to
amirm@. They also have a lot of data about DACs that you may found
useful. There is little correlation between price and quality in the
audio world and this data gives some objective criteria to decide
what to buy. You can
support ASR.
They provide a database of speaker measurements (manual)
Some scientific papers I have used:
Metrics for Constant Directivity (abstract,
paper,
poster) Authors: Sridhar, R., Tylka, J. G., Choueiri, E. Y.;
Publication: 140th Convention of the Audio Engineering
Society (AES 140) ; Date: May 26, 2016
A Database of Loudspeaker Polar Radiation Measurements (abstract, )
On the Calculation of Full and Partial Directivity Indices
(abstract); Authors: Tylka, J. G., Choueiri, E. Y.; Publication:
3D3A Lab Technical Report #1; Date: November 16, 2014
speakerdata2034
is a blog with a collection of spinorama from various sources. Index
of measurements is
available.
A number of companies provides data usually in the form of a GLL
file which is format developped by
AFMG. This format is very often used for live sound or tour speakers.
Here is a partial list of vendors providing data:
Standard Method of Measurement for In-Home Loudspeakers is available
for free at CTA
A Multiple Regression Model for Predicting Loudspeaker Preference
Using Objective Measurements: Part II - Development of the Model by
Sean E. Olive, AES Fellow. Convention paper 6190 from the
AES.
Farina, A. “Simultaneous Measurement of Impulse Response and
Distortion with a Swept-Sine Technique,” Presented at the AES 108th
Convention, Feb. 2000.
Hatziantoniou, P. D. and Mourjopoulos, J. N. “Generalized
Fractional-Octave Smoothing of Audio and Acoustic Responses,” J.
Audio Eng. Soc., 48(4):259‐280, 2000.
A. Mouchtaris, P. Reveliotis and C. Kyriakakis, "Inverse filter
design for immersive audio rendering over loudspeakers," in IEEE
Transactions on Multimedia, vol. 2, no. 2, pp. 77-87, June 2000,
doi: 10.1109/6046.845012.
M. A. Poletti and P. D. Teal, "A Superfast Toeplitz Matrix Inversion
Method for Single- and Multi-Channel Inverse Filters and Its
Application to Room Equalization," in IEEE/ACM Transactions on
Audio, Speech, and Language Processing, vol. 29, pp. 3144-3157,
2021, doi: 10.1109/TASLP.2021.3120650.
