Those are all good books, thanks for the advice. But I've searched for EBP in all 3 books and didn't find anything.

Yes I've read the paper of Small, but it doesn't explain exactly about the meaning of EBP.

So, I'm still searching for an answer.

In electronics, you've got this thing called "gain bandwidth product". See https://en.wikipedia.org/wiki/Gain%E2%80%93bandwidth_product. Basically, when you apply negative feedback to an amplifier, the gain drops but the bandwidth rises (it also gets more linear, but that's another story).

Since gain and bandwidth are in trade-off with each other (you can increase one by sacrificing the other), a figure of merit is desired which describes how good your gain is for a given bandwidth. This lets you compare different amplifier types before you've even decided how much feedback to use.

It would appear, although I can't prove this, that Small et all have borrowed from this idea in their definition of an efficiency bandwidth product (the EBP of the topic name). One important difference, however, is that amplification and feedback are not involved at all with EBP.

Instead, think about the throat of a horn flare. This will be an idealised horn flare whose throat loads the driver restively. This acoustic resistance adds on to the actual mechanical resistance of the driver as well as the effective resistance resulting from the electrical circuit seen through the motor (=magnet/voice coil system). As such the driver appears to be more damped than it would be in free air. If you now progressively narrow the throat, you'll increase this total acoustic resistance further and increase the damping further.

One effect of increasing the damping in this way is to increase the frequency at which resistance-limited behaviour gives way to mass-limited behaviour. If (and this is an assumption) you reckon that entering mass-limited behavior will cause a roll-off, then this frequency is at least in the ball-park of your system upper frequency limit, aka bandwidth.

Another effect of increasing the damping in this way is to change the impedance ratio between driver and horn throat, either for better or worse. If you keep increasing the throat resistance, you'll go past a perfect match and the impedance match will then only get worse, and then the efficiency will fall.

At least for throat resistances somewhat higher than driver source resistance, we discover a straightforward trade-off between efficiency (=how much noise can I make for given power) and bandwidth (=how high up can I play before mass-limited mode kicks in). The trade-off must have reminded Small of the gain-bandwith product which was established theory in electronics at the time.

The actual EBP (our figure of merit) is then defined as the transition from resistive to mass limited behaviour for the driver without any horn throat present. Thus, if the efficiency to bandwidth tradeoff is important to you, you'll look for high EBPs.

As a side note, there was a time when people were very worried about this tradeoff - it was central to the design of old public address systems for example. Now we can all buy comps that give 30% efficiency up to around 20KHz, and even if we do have efficiency/bandwidth problems (eg because we scaled them up in size to get more power, mass rose and EBP dropped) we can simply "pop in a supertweeter".

EBP still matters for low-mid and bass however. Low-mid horns are driven by cone drivers with EBPs usually under 500Hz. Getting up to say 3KHz may require "heavy" loading of the driver which will pull down sensitivity. However, the problem is only one of many that must be solved when designing such horns, and may not always be the most critical.

As for bass, port/cavity resonators (aka Helmholtz resonators) get hard to design when their resonance is near or above the EBP, when the driver is in mass-limited mode and therefore behaving as a mass-like source. It's much easier to control these resonant systems in the velocity-limited mode below EBP where the driver looks (to the cabinet) like a resistive source.

Hence the figure of 100Hz for ported boxes. This figure is assuming you want to cut off somewhere typical like 30-50Hz. EBP of 100+ ensures the reflex chamber and port can in principle be tuned in such as way as to obtain the bass extension. OTOH if not using a port, and you want a 40Hz -3dB cutoff, and EBP is over 100Hz, you've got a problem - what Fc and Qtc to use? I'll leave it to you to figure out why that won't work out well (hint: for a sealed box, EBP is roughly Fc/Qtc, to a first order approximation).

On a final note, EBP is also crucial for bandpass designs: similar to the above, multi chamber resonant enclosures also struggle to be controlled by the driver above EBP (and even near it). So if a bandpass bass system is going to stretch up to 150Hz+ (and it may need to in order to properly meet tops) the EBP should be >=150Hz, and this can rule out otherwise suitable looking drivers. The difference is, with bandpasses you don't get to "buy" the bandwidth you need by sacrificing efficiency - you just simply can't go above EBP. Well, not by much, unless you're really good