Frequency Response Specs

Going up, down and something in the middle

The frequency response is all about how low the bass goes, how high the treble goes, and what happens in between!

Often you’ll see it quoted as: 50Hz – 20KHz.  Sadly without also quoting the level tolerance this figure is useless.

Charts are best

Here’s the frequency response of a SoundBucket:

This chart is a typical, frequency response chart: the frequency goes left to right, logarithmically, and the level goes bottom to top, in dBs.  Because a dB is also a logarithmic measurement, they are spaced equally.  The reason the frequency is displayed like this is because our ears work in a similar fashion.  There are only 100 Hz between 100 and 200Hz, but musically this has the same weight as the 10,000Hz between 10,000 and 20,000Hz.

So, looking at the chart,  we can see that the response is essentially flat, but has wobbles from about 1000Hz (1KHz) upwards and the bass rolls off at the lower frequencies.  This is normal – bass can never be flat to DC, and the way that you measure the speaker always gives you more detail and/or noise at the high frequencies.  You can also see that the curve has been smoothed with a 1/24th octave filter.  This lets you see fine detail.  A lot of curves are shown with 1/3rd actave filtering, which removes all detail and is very flattering.

It’s generally accepted that the flatter the curve the more tonally accurate the loudspeaker will sound.  This makes sense: we can boost the bass to sound impressive for dance music, but if we then played a ballad, the voice would sound really boomy and wrong.  A speaker with a flat response will sound good with any/ every type of music.  The SoundBucket rolls gently downwards because it has an omni response, we’ll talk about this another time.

Now to measure the flatness, we measure how much the level deviates from the ideal flat line.  The best speakers in the world have a wide frequency response that only deviates by +/- 1dB ( 1dB louder, 1 dB quieter).  It’s again generally accepted that they need to do this over a frequency range of  at least 170Hz to 15KHz, which is where most of the musical information lies.  The SoundBuckets as you can see produce an exceptional response of 100Hz to 18KHz +/-2dB.

Response without tolerances

So how can we display the figures without a chart? Taking the SoundBucket chart above here’s some possibilities, all true.

The SoundBuckets frequency response is:

  • 30-20,000Hz
  • 100 – 20KHz, +/-2dB, ref 1KHz
  • 85 – 20,000Hz, +/-3dB, ref 1KHz
  • 80 – 20,000Hz, +/- 3dB
  • 60 – 20,000Hz, +/- 5dB
  • 30-20,000Hz, +/-10dB

You can see how misleading these figures can be!  What you also can’t see is what’s happening between the frequencies mentioned.  There could be massive peaks and troughs all over the response – you’d hear these as unpleasant colouration. But as long as those peaks and troughs weren’t bigger than the tolerance, you wouldn’t be lying!

Also it’s helpful to have a 0dB reference. The last set of figures above demonstrates this – 30Hz is 20dB less than the peak at 150Hz, but its 15dB less than the average level.

Finally, most portable speakers use digital loudness compensation.  What this means is that they boost the bass at low volumes to make it sound bassier.  This is OK, except that when the volume goes up, they can’t cope, so they reduce the amount of bass they produce.  At a low listening level they may boast a LF response of 60Hz, but at volume, the response  could change to a high 150Hz!

So how do you read response figures without charts?

  1. If there is no tolerance quoted, ignore it completely. It’s a useless measurement that does not tell you anything about the speaker
  2. The industry standard is a response quoted with +/- 3dB tolerance levels.  Ignore tolerances larger than this as they can hide nasty problems in the midrange.
  3. Ideally look for a reference point – 1KHz is typical, or you could be describing a curve, not a line.
  4. Look for a reference level for the reference point – 80dB at 1M is common.  If not the speaker could have a DSP bass cheat, and will just lose bass as it gets louder.
  5. The tighter the tolerance over the wider the frequency range the better

Finally it’s often useful to quote a figure when the bass drops below 10dB of the rest of the response.  This represents half the loudness, and it’s a good indicator of how low you can hear bass notes.  Most portable speakers drop off really quickly ( 4th order filter), but the SoundBuckets drop off half as fast (2nd order filter). From the above, the SoundBucket figure is about 50Hz. This should be shown as: LF cutoff: -10dB @ 50Hz (ideally referenced to 1KHz).

Last word

Not all charts are the same.  Speakers can be measured differently, the responses can be filtered to make them look smoother than they really are, and (my favourite) the vertical scale can be compressed.  A 10dB vertical scale is the standard, but beware no scale or 20dB scales – the responses can look amazing!