P3ESR Bass Measurements (misunderstanding?)

[Art K]

I found this review of the P3ES-2 and it mentions the bass response and drop off more specifically than I recall having seen before. Is the P3ESR that good or how does the bass response with the P3ESR differ/compare? Specific numbers like the ones in this article would be appreciated. Thank you.

http://www.stereotimes.com/speak061105.shtml

{Moderator's comment: Maybe you misunderstand the specification. Bass does not *suddenly* stop being reproduced at a specified frequency. Bass is correctly specified as being 'so many dBs down at the point that a certain frequency has been reached'. That specified frequency is higher up the musical scale with a small diaphragm or/and small box.}

[A.S.]

{Moderator's comment: You misunderstand the specification. Bass does not *suddenly* stop being reproduced at a specified frequency....

Agreed. Just to give you an impression of how the low-end actually behaves, here is a frequency plot of just that.

As you can see and just as you would expect, bass gradually (and predictably) rolls off with lowering frequency and diaphragm size. That is intuitive. Were that not so, we'd all be perfectly satisfied with the bass generated by 1" PC speakers!

What's really as or more interesting is the amount of power (or subjective weight) that can be generated at low frequency. That is not shown by a curve like this. That is a much more complex matter and really needs a listening test. Even in-ear headphones can produce loud, deep bass when used in-ear as designed. But remove them from the ear and even with turned-up volume there is no bass because there is no power ability at all.

Specifications are so extremely difficult to interpret it's no wonder most audio magazines don't even attempt to present measurements.

>

[Art K]

Thank you for the replies. What I'm trying to figure out is whether the P3ESR has as much energy or volume as the P3ES2 below 75hz? Does the P3ESR roll of as gradually? How many db down is it at below 75Hz (60hz, 50hz, 40hz and below). You give the spec of 75hz -3db.

{Moderator's comment: what have many users reported here after listening?}

[A.S.]

Thank you for the replies. What I'm trying to figure out is whether the P3ESR has as much energy or volume as the P3ES2 below 75hz? Does the P3ESR roll of as gradually? How many db down is it at below 75Hz (60hz, 50hz, 40hz and below). You give the spec of 75hz -3db.

The LS3/5a, P3 and P3ESR have very similar bass roll-offs as they are the same size cabinets with the same size drivers so you wouldn't expect much difference between them.

I think if you read back over my initial notes here about the design of the P3ESR I mentioned that I had three objectives in designing the P3ESR bass unit. That is, to increase the overall impedance of the P3ESR above that of the earlier P3(ES) to sidestep the occasionally reported power-hungry demands of the under four ohm P3(ES) and simultaneously to at least match the sensitivity of the LS3/5a. And to at least match the LS3/5a bass performance - the yardstick in mini-monitors.

When discussing roll-offs, it's not just an matter of quoting a simple number, not for a complex system like the P3-LS3/5a type mini monitor. For example, we could say that all three have a -3dB point at a certain frequency. But relative to what frequency? That requires a very careful look at the overall response shape below, say, 500Hz, and some judgement. Also, not quoted by that simple number is what happens down at, say, 40Hz. Theoretically for example, the speaker with the heaviest cone/coil (the P3ES) could produce more bass at 20Hz where a big, heavy slow cone is ideal. But that's ridiculous - how can we seriously expect a shoebox to produce 20Hz at a useful loudness even if we can measure some output at that low frequency in the lab under controlled conditions for a fraction of a second. What matters in practice is that we can we produce bass with adequate loudness without clipping - avoiding driving the tiny 5" unit beyond its safe operating limit - and giving the listener a full, rich warm listening experience to music not tones. That's far beyond one-dimensional dB figures.

To make the comparison even more difficult the sensitivity of the three speakers is not exactly the same and with very different amplifier power demands. The P3ESR and LS3/5a are quite similar. The earlier P3(ES) draws a lot more current from the amp due to its much lower impedance.

Frankly, if I was asked to critique mini-monitors I'd completely ignore the confusing and difficult to interpret bass specifications and actually listen for myself. The influence of the listening room will dominate the perceived bass of all speaker systems and introduce 'errors' of perhaps +/- 10dB in the bottom end response.

[A.S.]

As you may recall, I started on the design of the P3ESR woofer in 2005. The speaker was launched in June 2009. Admittedly, during this four year development cycle I had to undertake the M40 > M40.1 redesign, but even so, there were many aspects of the P3ESR woofer which needed careful consideration. And that takes time.

So I thought I'd have a look through my log book to see if there are examples of the sort of issues that a speaker designer - specifically a drive unit designer - has to consider especially in the lower frequencies. As with everything relating to loudspeaker performance, the outcome is at best a compromise. There are no ultimates. There is no perfect speaker. All technical solutions depend upon how the designer perceives his customer's requirements.

To illustrate some of the issues, attached are some graphs from 2005; early days and comparative measurements of woofer concepts.

Graph A:

This shows a number of different measurements of prototype woofers in a small box, without a crossover or tweeter. As I recall, it took me a day to make the woofer, mount it in the cabinet, make a frequency response measurement, store the result, remove the woofer from the cabinet, adjust its performance, reassemble into the cabinet, remeasure and repeat this several times. There are three things to note from this graph.

1) There is no absolute precision in the frequency response in the bass region (below, say 100Hz). Even the anechoic chamber is not perfectly absorptive at those low frequency, long wavelengths so there are lumps and bumps in the curves. The black curve seems especially irregular around 40Hz. Is this a feature of the measuring environment? Or the drive unit? Or the test signal? Or the drive level? Or temperature? Or some other factor? Truly cannot say with certainty.

2) If you were the marketing department pressing the speaker designer to supply a "-3dB" figure for the bass roll-off in the sales brochure, what figure would you give? What would be arguably the most marketable and how could you both bend that figure to have a degree of truth (in the right circumstances) without the designer threatening to resign?

3) Do you see that all the roll-off curves converge at about 70Hz downwards? On the face of it, you'd maybe say that these speakers had a similar low-end frequency response below 70Hz. And that would be true - the way this data is presented. That's because the curves are normalised to 60Hz or so. But look again and you'll see that that there are other perhaps less flattering interpretations that can be teased out. Mentally take the black curve (the loudest curve at 120Hz where the red cursor line is) and slide it down vertically at that frequency so that it is no louder than, say, the orange curve at 120Hz. Ummmm. Now what we see is that the roll-off of this specimen is far faster than the orange curve. So by presenting the same facts in a different way we have taken the black-curve specimen from apparently having the highest sensitivity (surely the best marketing spec.) to having by far the weakest bass.

Graph B:

Here we try and sidestep the vagaries of the anechoic chamber's performance at very low frequencies by creating a computer model of how a piston of the same mass, compliance and diameter would behave if set into motion at low frequencies. The result we obtain solely depends upon the data we feed into the simulator. But that's another big set of non-absolutes. What is the actual sound pressure generating area (or diameter) of the drive unit? Can we just measure that with a ruler? Do we include the surround? The dust cap? And how do we input accurate information about the springiness or compliance of the driver? How do we measure that without disassembling and destroying the driver? And as with all simulators, garbage in, garbage out.

This curve shows computer model predictions of frequency response that I created in 2005 for the two generations of LS3/5a woofers, the SP1003 and SP1228*. When you consider the number of variables, these 'first pass' models are quite close to reality - I improved them later. Even with the finest models, once the speaker is in your room (or even, as we've seen in Graph A the anechoic chamber) the response is really erratic at low frequencies. We here at Harbeth are unique at this scale of operation for designing and making our own bass units where we can 'bend' and optimise their technical parameters. Almost everyone else buys woofers with a pre-defined specification from a catalogue, mainly from far-off suppliers.

Conclusion: there are lies, damned lies and speaker measurements! What matters is listening.

*All Harbeth LS3/5a used the later KEF B110 SP1228 which we made for KEF as a spare part.

>

[Art K]

Excellent! Thank you, Alan.