I would add to that, reasonably constant directivity above 1kHz, with 90-ish deg beamwidth in the horizontal plane. HD products restricted to the audible range.
I'm not well versed in horn-loaded designs, but I believe no such combination exists that's commercially available? For all I know, the limitations imposed by non-linear acoustics may prevent such an outcome...
Thanks.
I'm not well versed in horn-loaded designs, but I believe no such combination exists that's commercially available? For all I know, the limitations imposed by non-linear acoustics may prevent such an outcome...
Thanks.
Last edited:
Where come these specs from?
Just had a look into the 1" drivers measured for my cinema project - I get about 2% THD at 110dBSpl. But that's mainly 2nd harmonic, 3rd harmonic is down at -60dB. (BMS and B&C ring radiator drivers).
But even the Bliesma T34A has about 1-2% at 105dBSpl - so this is a pretty hard task. Maybe a big AMT can do it (with other downfalls of course).
I really always have a close look at THD cause there is a lot of information in there - but honestly I'm totally find with 1-2% mainly 2nd harmonics at these levels.
Just had a look into the 1" drivers measured for my cinema project - I get about 2% THD at 110dBSpl. But that's mainly 2nd harmonic, 3rd harmonic is down at -60dB. (BMS and B&C ring radiator drivers).
But even the Bliesma T34A has about 1-2% at 105dBSpl - so this is a pretty hard task. Maybe a big AMT can do it (with other downfalls of course).
I really always have a close look at THD cause there is a lot of information in there - but honestly I'm totally find with 1-2% mainly 2nd harmonics at these levels.
The answer is no. But the BMS 4599HE comes close at -45dB (0.57%) at 105dB.
Attachments
What Dr Geddes said about THD coralated to a good perception ? I remember there is a polemic and one should look at elswhere : storage energy, etc.
Is there a concensus than odd harmonics should be at minima -55 dB for 3H ; -70 dB for H5, etc ?
Is there a concensus than odd harmonics should be at minima -55 dB for 3H ; -70 dB for H5, etc ?
It is reasonably straightforward these days to use DSP to cancel a substantial part of the distortion due to nonlinear compression/expansion. Not sure if there is enough interest/demand for commercial products but it is DIYable.
I never experimented with added harmonics to detect influences and listening levels - but in music it's for sure higher as -55dB for H3. But of course it would be a good idea to simply stay below that just to be save ;-)
But I have experience with BMS 2"/1" coax driver neodym version - cause I built my main PA speakers with them. When you crank them you get plenty of 2nd harmonics. They never sound harsch or resonant as most 2" drivers but they add some "sweetness", hide finest details. This is no problem with recorded music - but it takes away a little the direct "connection" between musicians, mixer and output. Mids & highs are more polished as real life. Some colleagues don't like that driver for that purpose - I still prefere it over all the other 2" drivers of that time (that's a few years in the past ;-)) cause of the detailed high frequencies (I mixed a lot of percussion ensembles - great for that).
But I have experience with BMS 2"/1" coax driver neodym version - cause I built my main PA speakers with them. When you crank them you get plenty of 2nd harmonics. They never sound harsch or resonant as most 2" drivers but they add some "sweetness", hide finest details. This is no problem with recorded music - but it takes away a little the direct "connection" between musicians, mixer and output. Mids & highs are more polished as real life. Some colleagues don't like that driver for that purpose - I still prefere it over all the other 2" drivers of that time (that's a few years in the past ;-)) cause of the detailed high frequencies (I mixed a lot of percussion ensembles - great for that).
Intersting! Is it that you feel it sounded more "nice" or is it actually a more faithful reproduction of e.g. a congas?
//
//
I have been thinking about this since I read cd distortion has a mathematical model that makes it predictable (Beranek). Did someone operationalize this knowledge into something useable? Would love to find out more.
It has been doable for many decades in terns of maths and signal processing. I don't know about commercial implementations because deep horns are not a field of interest. I would expect there to be a few academic publications given the nature of the topic but I haven't looked. The coefficients may need to be horn specific though because the distortion will change during propagation when the SPL is high. Whether enough to be significant I haven't considered. Looks like a nice DIY project for someone interested in horns and signal processing.
Congas are not that hard to reproduce if you have enough dynamics cause you don't touch the ends of the spectrum. So I don't think I would hear a big difference with them. West african djembe would be more demanding, esp with the rattle "ears" (kessing/sesse) added. (and they can mess up your whole stage cause they are so freaking loud, especially when they are right behind the balafon 🤓)
Maybe for a HiFi comparison - it reminds me of a (good) silk tweeter vs a hard dome. Or a silk midrange dome vs a hard membrane midrange (which I just did a comparison a few weeks ago). You have all the music here and also most details - but a strong crash is a little "softer" as the real metal thing hit with a hard stick. Dense music (most stuff with distorted electrical guitar get's dense pretty quick) is shown as a "complete" thing but single signals are not so clear seperated. For some music that's nice and live for sure better as a harsch 2" with breakup >7kHz (like the old B&C drivers) - but not 100% true.
Don't get me wrong - at home levels I don't hear this effect. That's only when you really push these drivers for a bigger audience (I would be 20-30m away from the speakers then). A little "rounding" can be a very welcome thing with live signals but it's not everyones cup of tea.
But maybe enough vapouring about subjective sound and back to some nice and easy THD measurements ... ;-)
Andy,
If you have any information of any implementations of "maths and signal processing" showing reduction in harmonic distortion to levels below what current (or past) high frequency compression drivers have, please post them !
David Gunness has decades in the field of commercial implementations of signal processing specifically addressing correctable problems in horn loudspeakers. He has published many papers and patents, including the October 2005 "Improving Loudspeaker Transient Response with Digital Signal Processing"(PDF). Convention Paper. Audio Engineering Society.
Although those processes, dubbed "Gunness Focusing" during his time at EAW address and largely correct four different problems, harmonic distortion is not among those.
The four correctable issues are "linear, time-invariant" (LTI) mechanisms, which are arguably greater in terms of sonic degradation than harmonic distortion:
1) All phase plug designs produce significant smearing of the transient response.
A significant fraction of the sound energy arriving at a phase plug opening will either continue past it or reflect back from it; in either case arriving later at other phase plug slots where the sound is divided again, ad infinitum.
A carefully constructed filter falls short of perfectly correcting the axial response, but it improves it significantly, while also significantly improving the off axis response.
2) A second loudspeaker behavior, which yields well to digital preconditioning, is horn resonance.
3) A third behavior, which also yields to digital preconditioning, is cone resonance.
4) A fourth behavior is non-linear phase response of the summed crossover.
The essence of the filter to correct that defect is the high frequencies are delayed relative to the low frequencies, which counteracts the minimum phase crossover’s effect of delaying the low frequencies relative to the highs. This requires an added high frequency latency of the length of the FIR filter.
All four combined can indeed make a compression driver based loudspeaker sound much better.
Art
Great conversation so far - thanks for the many enlightening comments. Some further reactions/thoughts:
1. From whence -50dB? Historically, the threshold beneath which I'm pretty sure I can detect no sonic degradation. 108 dB at 1m isn't my real target; it's more like 98-99-ish dB at 3m. I also recognize that at higher SPLs the distortion of the ear-brain mechanism plays an increasing factor.
2. -45dB at 105dB (at 1m?) is still quite good, and if I have to tolerate HD, best the second harmonic, but 2nd-order HD can still sound quite bad under certain circumstances, so I still include it under the -50dB THD target.
3. DSP-based distortion cancellation - somewhat like what Klippel is doing - very interesting. But it's news to me that this doable as "DIY"?
4. I'm aware that there's more to distortion than just HD, but since it's the most frequently cited spec, and because IMD tends to be related, it was the handiest to quote in my subject line.
5. Weltersys - very interesting. Quite sophisticated. But is this "DIY-able" as andy19191 suggests?
Thanks.
1. From whence -50dB? Historically, the threshold beneath which I'm pretty sure I can detect no sonic degradation. 108 dB at 1m isn't my real target; it's more like 98-99-ish dB at 3m. I also recognize that at higher SPLs the distortion of the ear-brain mechanism plays an increasing factor.
2. -45dB at 105dB (at 1m?) is still quite good, and if I have to tolerate HD, best the second harmonic, but 2nd-order HD can still sound quite bad under certain circumstances, so I still include it under the -50dB THD target.
3. DSP-based distortion cancellation - somewhat like what Klippel is doing - very interesting. But it's news to me that this doable as "DIY"?
4. I'm aware that there's more to distortion than just HD, but since it's the most frequently cited spec, and because IMD tends to be related, it was the handiest to quote in my subject line.
5. Weltersys - very interesting. Quite sophisticated. But is this "DIY-able" as andy19191 suggests?
Thanks.
Meanwhile, the best is to increase diag Sd at iso size throat output to reduce HD. For instance big EV compression, see Weltersys compressions test.
"...Weltersys compressions test" where do I find that?
In pondering this a bit further, let me limit the consideration to non-linear mechanisms. In fact, that's the reason why I cited THD at a specific SPL at the outset.
Another reason why I cited -50dB btw is that this is achievable - with some difficulty - by a direct radiator, or more likely, an array of direct radiators. Adding a horn adds to the difficulty but hopefully not to the point of making it infeasible. (But let me not admit the possibility of a horn array. I think arrays of direct radiators are admissible because they tend to be areally much more compact.)
In pondering this a bit further, let me limit the consideration to non-linear mechanisms. In fact, that's the reason why I cited THD at a specific SPL at the outset.
Another reason why I cited -50dB btw is that this is achievable - with some difficulty - by a direct radiator, or more likely, an array of direct radiators. Adding a horn adds to the difficulty but hopefully not to the point of making it infeasible. (But let me not admit the possibility of a horn array. I think arrays of direct radiators are admissible because they tend to be areally much more compact.)
Absolutely doable as DIY and also "free", if you have a computer to spare -> CamillaDSP https://www.diyaudio.com/community/...ne-for-crossovers-room-correction-etc.349818/
//
//
I have never even considered using a compression driver and/or a deep horn for a speaker design (I probably should) and have no knowledge of what has or has not been implemented in the area. I understand the basics of acoustics and signal processing and my comment was solely based on this.
At low SPLs assuming sound creation and propagation is linear involves negligible levels of error/distortion. As SPL rises this becomes less the case but we have do have a reasonable model for the nonlinearity in assuming a polytropic compression process $pv^\gamma$ [can we write maths on this forum?]. This can be used to predict the distortion and hence cancel it by adding the inverse to the input signal. As mentioned above this sort of thing is in basic acoustic texts from the 40s and 50s (e.g. Beranek) and, no doubt, many horn focussed publications since.
A DIY project to implement this will only involve time and effort since the hardware involved is likely to be already owned by an active speaker DIYer but it will depend on knowing (or being interested in learning about) the physics and maths involved in acoustics and horns plus how to design and implement signal processing. Of course, if the SPLs at the throat of the horn are not high then this form of distortion won't be dominant making cancelling it fairly pointless.
With MUSIC harmonic distortion is way less noticeable - cause all the instruments and sounds have plenty of it 😉
It's already pretty good when you can detect -50dB with sine in the mid of the audio range (get's hard past 10kHz 🤓 ) - no way you hear it with music.
You should be able to get in this area with 3" mid dome and the 34mm Bliesma tweeter when you accept a little higer 2nd harmonic. And want wide dispersion instead of a horn.
Ok. So, LTI stuff - I basically have various means at my disposal to deal with that. It's the non-linear stuff where things get challenging. So let me return to a horn/driver combo with nothing beyond LTI correction that gets me to -50dB THD. I suspect that it may be beyond reach because even direct radiators would be highly challenged to meet that goal.
Sorry - forgot to add - I want directivity control too, beyond what a certain size piston vibrating in a baffle would render per the wave equation. The horn may be a minus from a distortion-adding perspective, but it's a plus in terms of reducing distortion from the transducer given the overall SPL target and the horn's effect to increase efficiency.
Diyiggy was probably thinking of this:
High Frequency Compression Driver Evaluation
I think the non-linearity of air by itself may make -50dB THD at 108 dB at one meter at 1kHz fairly close to "as good as it gets" for a compression driver, as peak SPL at the diaphragm/phaseplug entrance would exceed 150dB at that level for a 1.75" 1" exit driver on a 90x45 degree horn. The pressure changes within the air medium cause the wave energy to transfer to higher harmonics- can't change that with any compression driver design or pre-conditioning filter.
At any rate, I can't think of a listening situation where I could have been able to detect -50 levels of THD, even back when my HF hearing was +50dB above what it is now ;^).
That said, it only requires about one electrical watt to generate 108 dB at one meter at 1kHz, while peaks in the 100 watt range may be common in PA use- you may not want to know the distortion levels that are typical, but the attached JBL Tech notes Vol 1, Number 8: "Characteristics of High-Frequency Compression DriversPDF shows Beranek's distortion calculations and how well they agree with measurements ;^)
Correcting "linear, time-invariant" (LTI) mechanisms are now possible with relatively inexpensive FIR (Finite Impulse Response) processors, and many "DIY" approaches have been done that emulate what David Gunness has done.
I'd suggest working on what can be corrected, and not worrying about what can't .
Art