cool audio stuff and recorded music last updated 26 January 2009
I use the Edgar Monolith bass horn. I built a stereo pair with JBL 2220J drivers; their Qes=0.18 is ideal for horn applications, it allows for a higher frequency range. I sand loaded the side panels and mass loaded the top panels with marble slabs.
Edgar claims it is the best 1/8 size bass horn he has designed. However, do not use it with wooden floors: the down-firing exit will play bass through the floor and out of the room! My floor is concrete and I get solid output to 40Hz as per the horn design, then rapid rolloff. However I have equalised the output flat to 20Hz. I also use angled concrete slabs as reflectors into the room, to prevent back-reflections from the floor up into the horn.
I thought I would need a subwoofer to complement the bass horns for deepest bass 20-40Hz but I was completely wrong. A big feature of Edgar bass horns is that they are uncompromised: they are not truncated (cut short to save space, a very common practice in commercial bass horns, also in diy project designs if truth be told), so the flare is full theoretical length and the mouth is full theoretical area for a 40Hz 1/8th space corner horn. Combined with a proper tuned back chamber volume individually calibrated for each drive unit in service, and this horn maintains low distortion, low cone excursion, and full efficiency at the 40Hz cutoff frequency! Efficiency is 105dB per watt and the drivers are rated 200W RMS or 600W peak. That equates to 128dB continuous RMS per channel at 40Hz; try that with your typical interior home subwoofer and see what happens. The Shiva subwoofer I built, tuned for 16-50Hz and 110dB output at 20Hz, is clearly superfluous, so I am now looking for a new application for it around the home.
Transcript of an email I sent a friend regarding the Karlson Speaker (a bass unit from the 50's with a cult following), and concluding with some comments on good bass.
I have been using the JBL M209-8 200mm pro audio midrange on an MDF flat panel. It is a very good driver with rated 96dB sensitivity but actual is higher, perhaps due to beaming at the highest frequencies. This is a genuine midrange driver and not a "music" driver like the other JBL M-series drivers. JBL use it in their Venue-series permanent installation speakers with a 220Hz horn: see the VS2210. It is suited for use with or without a horn. With no crossover it runs from 240Hz (limited by the open baffle's size) to a little over 2kHz. I have been running it crossed over to the Monolith at 250Hz, and with no upper crossover, allowing natural rolloff.
I had intended to build a huge 150Hz horn on the front of this driver, almost 1m in diameter. However, I could not find a solution to the problem of tweeter placement. I am not prepared to compromise the engineering of my system with a 600mm or 700mm axial spacing between the tweeter and midrange, even though Avant-Garde insult their customers with such poor engineering in a US$39,000 product.
I also found that even this huge horn was not going to provide the low frequency dispersion control that I require. Once I had decided not to use a horn, the JBL driver was no longer optimal for my design.
Transcript of an email discussion with a friend regarding the tractrix horn flare, which is mistakenly believed to be 'perfect' in the sense that it conforms to the natural expansion of a spherical wave front and hence has no reflections off its walls.
In 2007 I am upgrading the midrange. I am continuing to use a high output pro audio cone driver on a flat panel. For my rationale see Loudspeaker System Concept, below.
The flat baffle is being upgraded from MDF to a corian panel with slightly larger dimensions (sufficient for 200Hz low frequency) and with a constrained layer damping system for excellent control of resonance. The CLD comprises a non-bituminous damping compound 1mm thick sandwiched between the back of the corian panel and a 1mm aluminium sheet. It is important that the aluminium sheet is thin, so as not to increase the stiffness of the composite sandwich, which would only increase its resonance and reduce the effectiveness of its damping layer.
The JBL driver is being replaced with a contemporary 250mm cone from 18 Sound, the 10NDA520. This is perhaps the best driver available for my design, with its high efficiency, low distortion, and high excursion (for a pro audio midrange). This driver has a number of features that contribute to its very low distortion, including the neodymium magnets and the patented active impedance control technology.
The new midrange will operate 200Hz to 1000Hz, with a constant beamwidth of 100 degrees.
I have been using the JBL 2204H horn tweeter with a 2.7uF series capacitor, for a 7.4kHz first order crossover. This leaves a hollow in the frequency response between 7kHz and 2.5kHz where the midrange JBL cone dies naturally. I have equalised the hollow flat. This is not a good design and was always going to be temporary.
In 2007 I am upgrading the treble to a two-driver array. The low treble will be handled by an outstanding 1" compression driver from 18 Sound, the top of the range NSD1095N, on the 18 Sound XT1086 constant directivity horn, between 1kHz and 10kHz. High treble will be handed over to the trusty 2404H above 10kHz, where it performs best.
The NSD1095N employs a new, patented phase plug technology that reduces distortion and frequency response aberrations inherent to the traditional concentric phase plug designs. More details here, click the 3P java link.
The XT1086 is a constant directivity horn with a smooth flare. There is no discontinuity or dispersion slot, as found in bi-radial horns and similar designs, which causes back-pressure reflections.
The NSD1095N/XT1086 combination delivers less than 0.1% third harmonic distortion over its entire operating range. I consider that inaudible, and that is at 1 watt input power which equates to 110dB SPL. Compare that to the distortion of even the most esoteric home audio tweeters at 110dB SPL. Furthermore, given that speaker distortion decreases monotonically with input power, the third harmonic will be well below 0.1% at even the loudest listening level peaks.
At frequencies above 12kHz, the NSD1095N/XT1086 combination gradually increases directivity, i.e. beamwidth decreases below 80 degrees horizontally and 60 degrees vertically. According to the best theory I have access to, such an increase in directivity is ideal. I may not need the JBL 2404H in this case, with its 100x100 degree dispersion up to 20kHz. I will consider the question of whether to use the 2404H above 10kHz as a personal tuning preference in making a final decision.
The above speaker system resulted from my search for the best loudspeaker to achieve my objectives, namely, low distortion and a controlled acoustic result (direct + reflected) at the listener's ear so that the listener's experience is close to the experience of being in place of the recording microphone. The system need only exceed the known psychoacoustic parameters of the human ear.
- Sidebar: I don't subscribe to the "infinite subtlety" notion of auditory perception. The ear's ability to respond to audio is well understood, and so is the mind's ability to distinguish changes in the ear's behaviour. Our strong suits are localisation and dynamic range. But there is a tendency amongst audio enthusiasts to overrate our ability to discern bandwidth, distortion, phase errors, and changes in loudness. Despite enormous publicity for an experiment that proved to be invalid (irreproducible), there is no valid double blind experiment that supports the claim that ultrasonic frequencies (over 20kHz) somehow contribute to human perception. Sadly, a whole range of high-end over-20kHz supertweeters is now finding its way into loudspeakers based on a falsehood. Similarly, phase errors are not audible in music unless they create an amplitude error which is in itself audible. Therefore using equalisation to smooth out response errors will reduce the total system phase error, not increase it. An entire generation of audiophiles has been denied the benefits of good EQ because of falsehoods about phase errors, and coerced to use amplifiers devoid of EQ, which forces them to listen to massive total system phase errors that could cheaply and easily be reduced. I am pragmatic about these things.
I investigated a range of loudspeaker principles with an open mind: point source and line source, direct radiator and waveguide and horn, cone and dome and ribbon and electrostatic, transmission line and bass reflex and infinite baffle, monopole and dipole and omnipole. Here are my conclusions.
Low Distortion. For distortion to be low at all listening levels, I take that to include the 105dB standard used by THX. Not that I am into home theatre, but the sound levels I hear in movie theatres are the levels I want to reproduce in the home, effortlessly. However, even THX standards only require the 105dB level to be reproduced; it is not specific about distortion, and I want low distortion at 105dB.
The first challenge for the audiophile is even to find data on distortion of loudspeakers and drive units. There is plenty of data on frequency response (usually only published for anechoic chambers where home owners never use them, because in a normal listening room the response is horrific) and impedance data, but no distortion data. Why? Because speaker distortion levels are higher than the customer would like to see. If you find the distortion data for a speaker plotted against frequency, you will firstly note with displeasure how high it is below 100-200 Hz. If it is a legendary low-distortion speaker like the Quad electrostatic, midrange distortion may be a low 0.1%. For a 'normal good' speaker 0.5-1% is common.
But these figures are at one watt of input power (even though the distortion is measured at the most flattering frequency, but may be over 5% at some other frequency). For the Quad, one watt equates to an SPL of only 82dB! As you increase the power, the distortion increases monotonically for all speakers. 50w of input to the Quad is only 97dB, yet that nice 0.1% distortion is now several percent and she is on the point of arcing out (which I will call 100% distortion). This is clearly of no use for low distortion realistically loud sound for all forms of music.
The 105dB-with-low-distortion challenge quickly eliminates electrostatics, ribbon bass and mids, and practically any direct radiation bass driver or dome treble driver. I found that only a bass horn can keep distortion low below 100Hz, and only a horn or waveguide compression driver can handle 105dB treble, although only the best of these have low distortion. For example, my 18 Sound driver/waveguide combination with one watt of input produces less than 0.1% distortion from 1kHz to 20kHz. This is quite exceptional, and that one watt equates to 110dB of sound!
House curve: some advice on its use or non-use
I have about 8 years’ experience in equalisation of my home audio system, having owned and used both the current version and an earlier incarnation of Behringer’s digital equaliser, which has built-in pink noise RTA tools.
In my efforts to identify an appropriate target EQ curve for the in-room response to pink noise, I came across the X curve standard in pro audio, and have had discussions with local AES members, who have expertise in home audio, cinema and venue audio, and one fortunate member with crossover experience in both fields.
More recently I have had a look at the excellent efforts of this forum’s members to come to grips with the same question of the choice of target curve for use with REW. I particularly appreciated the efforts of Wayne A. Pflughaupt with the sticky note “House curve: What it is, why you need it, how to do it!”.
However, this is a topic that is rife with misunderstanding, because it is difficult to conceptualise (for me, at least, and I presume also for others like me), and some misunderstandings have crept into Wayne’s writings. I would not concern forum readers and fellow enthusiasts with a few minor points of correction, but if a misunderstanding leads to the wrong recommended course of action, i.e. the wrong choice of target curve, then I must speak my mind.
My experience in equalising to flat at the listening position with pink noise was exactly like Wayne’s early experience: searing treble and AWOL bass. Applying the X curve, or half or three quarters of the X curve, as a target brought things back to the realm of listenable, although more than half of X invariably meant loss of sparkle in the highest treble.
However, my bottom line is that the amount of HF cut you need will vary from recording to recording, and some correctly made recordings will produce correct sound in your room (and mine) with no treble attenuation – irrespective of the size of our respective rooms. I could not make that claim if Wayne’s statement were true, that “speakers sound brighter the closer you get to them. Therefore we must compensate with a tilted response curve that reduces the highs and emphasizes the lows” in the fifth paragraph under the heading “It’s all about the room”. I also cannot leave be the stated notion that the house curve is one where the listener hears all notes or tones at the same loudness, as stated by Wayne in his third paragraph under the heading “An easy way to determine the house curve you need”.
Rather than argue points, let me explain by way of an example.
Let us say an orchestra (or solo piano) is playing in a world class venue, and the recording mic is placed in the best seat in the house, say centre stage, 20 rows back. No EQ or compression is applied to the recording. When we play this back in our homes, what is the ideal target curve? It is flat, not X. Flat EQ will reproduce the same bass-to-treble balance in the listening seat as in the seat where the mic was placed. Playing the above recording through an X curved system will sound far too dead due to the attenuated treble. Our homes are not big enough to require an X curve on the basis of room size.
There are other reasons for the bass-shy sound in our homes with flat EQ. One is that the recording mic is usually very close to the instrument, which provides a completely different tonal balance to a mic in the audience seat. Another reason is the generic use of multibanded compression as part of the production process. I have heard side-by-side master edits of acoustic music with and without compression, and it isn’t the bass that gets boosted as a result of compression!
My conclusions? Yes, most recordings sound too sharp with flat EQ’d pink noise at the listener’s seat. No, it is not due to the X curve effect. Yes, applying the X curve helps with many (most) recordings, but only fortuitously and not with any accuracy; for example, the knee need not be at 2kHz and the slope need not be 3db/octave. After all, the sound errors are not being caused by the effect that the X curve correctly compensates. So feel free to experiment more freely with target curves, and recognise that different recordings need different EQ. So many recordings are badly done that a fixation on “one right setting” is only a means to bring suffering upon oneself.
To understand this issue properly, it is important to begin with how we perceive. It is a humbling lesson. Speaking from personal experience, it taught me to be less arrogant about my perceptions, particularly my sonic impressions.
to be continued....
A corollary on the audibility of resistors.