What are the pros & cons of high-efficiency horns?
- Thread starter manisandher
- Start date
Nothing you said after this supports the statement. But I don't think it's theory, I'd say there is more fact in the matter if we consider noise as well. Tube amps often sound louder even if SPL doesn't show it, and they have a little more distortion so that would correlate - but I think it has more to do with the consequences of using transformers on the output. And that would be linear distortion, not distortion as we typically think about it (phase, timing, T/HD).
What really does show is differences in noise. I've heard many systems where perceived dynamics was related to the amount of noise. The more noise removed from the AC, the less "dynamics" there were. Now this can only be true if the amplifier is having a lot of sag as a more pure AC wave actually can introduce slightly more sag. The trouble is people make these remarks regardless of volume. What I've also noticed is the ear piercing fatigue that is the "dynamics" in these cases. There are times when rather poor methods are used to clean up the noise, so I can't fault those scenarios, but in general it seems like dynamics are often associated with "loudness" feeling, and sharpness. "Loudness" also tends to sound sorta busy and blurry, and sharpness is pain without SPL... I don't call them dynamics, but people are often use to them so sometimes an album they liked sounds flat later with less noise, and thus is less dynamic. I think that once you've controlled a lot of noise and stuff you need to reevaluate what you thought was a dynamic record. There are some digital records of albums that have been tested for crest factor you can use, to find truly dynamic recordings. A lot of people will be surprised what actual SPL differences sound like.
Physics demonstrations need numbers - please tell us exactly how many dB's of peak compression you listen in a Wilson or Magico operated at 100 dB peak. I am not addressing show or demo levels, just the sound levels most people listen at home. My comments on loudness levels were based on recent data reported by people in this forum and experience at horn demos - always carried at louder levels than equivalent demos of box speakers. Perhaps your experience may differ.
And yes, the Dynaudio 1.8 sounded really boring. But operated as a domestic speaker the Dynaudio Special One (82 dB/W) paired with Cello electronics or my Sonus Faber Amator (86dB/W) with tubes, both using the same tweeter, sound really great!
Anyway, dynamic compression is a dangerous argument for analog lovers - tape and vinyl compress a lot more than digital, that is compression free. To the point that many people associated the preference for analogue to the natural compression of this media, feeling that the full dynamics of real events is not adequate to sound reproduction.
BTW,1 you forgot to address the 500W amplifiers of vinyl head cutters. They heat so much that sometimes they are cooled with water or even helium - using negative feedback to control them.
BTW2, the WAMM did not sound compressed at all, even when Peter McGrath wanted to play them a bit louder - just for fun, as he told. No one I know has ever complained that Alexandria's sound compressed.
I think we actually agree on some points. It seems you agree the drivers are the key. All high sensitivity drivers are on the right side of the power curve. But the best of all is high sensitivity combined with high power handling. This can definitely put off the onset of power compression.
https://www.linkedin.com/pulse/power-compression-vs-thermal-distortion-loudspeaker-alexander-wilson
This reference is really not a complete picture across the industry. Driver designers have been addressing power compression for decades. Where did he get the 35% distortion number? A drop in current flow is not a real metric that equates to a distortion percentage. What you are going to get is a measurable drop in SPL. Also why is an active crossover going to minimize power compression in a driver?
As a side note the 2216 woofer used in the JBL M2 has special low TCR VC wire that does not heat as quickly as conventional VC wire specifically to address power compression.
Rob:}
Last edited:
Driver compression really matters for how loud you listen, how big the room is, and how many drivers there are. A line array won’t exhibit the same amount even if the sensitivity is lower, because you’ve spread the thermal load across all the drivers. Similarly, midranges and tweeters don’t move a lot, and take smaller amounts of power.
Think of a driver like an 8ohm resistor. Flow 2w through it for say 91db, in a 87db sensitive speakers. How much heat are generating? Not much, and not 100% duty cycle. But if you play really bass heavy stuff, and you’re spending hundreds of watts, it’s going to become very real.
But the actual dynamic range for domes is significantly lower, they interact with everything around them a lot more, and they tend to have pretty different properties (including higher distortion).
I personally think the difference in driver properties plays an enormous roll. The QTS is very different, and across the board more well damped drivers sound closer or alike horns, which are also very damped. The difference is the horn can play louder with a wider range when QTS is low, and the non-horn also has lower xmax but with baffle step loss instead of gain (what horns do). There are some other factors like FS, to that statement, but in general it’s true that non-horns have little output anywhere near their FS, and horns might be concert level.
Think of a driver like an 8ohm resistor. Flow 2w through it for say 91db, in a 87db sensitive speakers. How much heat are generating? Not much, and not 100% duty cycle. But if you play really bass heavy stuff, and you’re spending hundreds of watts, it’s going to become very real.
But the actual dynamic range for domes is significantly lower, they interact with everything around them a lot more, and they tend to have pretty different properties (including higher distortion).
I personally think the difference in driver properties plays an enormous roll. The QTS is very different, and across the board more well damped drivers sound closer or alike horns, which are also very damped. The difference is the horn can play louder with a wider range when QTS is low, and the non-horn also has lower xmax but with baffle step loss instead of gain (what horns do). There are some other factors like FS, to that statement, but in general it’s true that non-horns have little output anywhere near their FS, and horns might be concert level.
I would concede that noise is a culprit as well. Tubes can sound louder for other reasons as well (look at the measurements made by Peter Van Willenswaard in a couple of Stereophile articles. https://www.stereophile.com/features/357/index.html )
"And that would be linear distortion, not distortion as we typically think about it (phase, timing, T/HD)."
In this discussion about compression a linear distortion is a distortion...not scaling the peaks in the music signal correctly is a distortion of the original waveform.
Yes, your observations regarding noise mirror mine but I would say that I have found that the cleaner the power the deeper into the music, more 3d etc. and more true dynamics come out. What you describe is the fake dynamics that I was referring to, which can definitely be affected by the electronics/power (in fact this is probably a larger culprit but I was trying to stay focused on topic of speakers here).
I have found with formats it is largely irrelevant because how many recordings are using the available range in the first place? Since the loudness wars many digital recordings are using only a fraction of the range. Older vinyl recordings might have a lower ultimate limit but they are pushed far closer to that limit. That being said, I generally (although I have a few notable exceptions) prefer large classical works on digital, finding them less compressed and more open, especially when it gets loud and/or really complex (usually the same thing). Small classical I give the edge to vinyl generally. The notable exception that comes immediately to mind is my 1965 Wagner set from Decca "The Valkyries" conducted by Solti. This sounds unreal dynamic but they took extra care in the LP production to minimize distortions...they got the whole range possible from vinyl on that set.
Don't know about other Dynaudios but I also had a pair of Madisound kits that used Dynaudio drivers and it was perhaps a bit better than what I described with the 1.8 but only a bit. I also have not liked any of the last 20 years of Confidence series, hearing the same issues. Only the Evidence was decent...but not exceptional in this regard.
As I said earlier, the big Wilsons are probably one of the very few dynamic speakers that I found to be relatively dynamic...especially the old X1s, which were a good 2-3db more sensitive than later ones. Smaller Wilsons not only don't have the dynamics of their big brother nor are they at least nice an flat FR like their competitors...lose/lose IMO. The biggest Focal is also not bad in this regard but I don't like the sound of the materials in their drives for some reason...or how they tune their crossovers (always a bit bright).
The speakers I sell, Dynamikks, are largely non-horn (only the tweeter), but they use large diameter, high sensitivity drivers, for the mid and woofer to try to "keep up" with the horn tweeter. They succeed pretty well and have big effortless sound (every bit as much as I have heard from a big Wilson) ...but it is not worlds apart other than the tweeter decision (I think Wilson's soft dome is a big mistake...but that's just me).
As the articles point out, the impact of thermal distortion is always there to some degree...long before the actual onset of power compression. Distortion in this case is also linear distortion (i.e. SPL drop) and not just non-linear distortion (although there is that as well).
At domestic levels though a few mA going through a coil will cause far less heating than half an amp or an amp. What I have not found addressed really well in the literature though is what happens when you have a 30db swell in an orchestral crescendo and how rapidly this heating occurs and what suppression of db that will result. I found one article some time ago (now can't find...wish I had saved it) that said they measured 7db of drop. I suspect that tweeters are more susceptible as they have much finer voice coils windings and then would heat up more quickly (depends also on the frequency range used). Many manufacturers use or used ferro fluid to help this issue but there are many reports of ferro fluid damaging the sound in other ways (mechanical resistance increases?). My horns only need a few mW to play at levels that are more than satisfactory...and they have large 3 inch voice coils, and they literally barely move and they are quite flat in response.
Another point that was addressed in some of the research I read is that since most drivers will react differently to thermal compression then it becomes an issue with crossover points, box tuning and other assymmetric issues in the overall design. The speaker is literally shifting around (soncially) compared to some fixed level points. That could result in a rather bizzarre sound under certain conditions I think.
JBL has a lot of knowledge in these areas of course; however, their focus is primarily pro sound where the coils can run quite hot for long periods of time and they have built the stuff so that the output won't droop over the course of a concert or disco . This has positive effects of course in the domestic setting where the drivers are barely breathing compared to many domestic speakers with tiny woofers, mids and tweeters of low sensitivity and not much thought about thermal effects. The hard part of pro drivers is getting them sound smooth but there are ones out there that are significantly better in this way and then you can get real high end refinement from them.
I was at a friends last weekend who has the Sonus Faber Olympica 3, which is a mid-size decent sounding speaker. At lower volumes it is clean, articulate and reasonably lively (but not exceptional this way). Woofers are 8 inch (2 of them) and mid is 4-5 inch and 1 inch dome tweeter (maybe 1 1/8 inch). He is running a Devialet something or other. At mid volumes the sound still is pretty clean but you can hear that the dynamic contrasts are already shrinking. Go a bit higher and the sound hardens and becomes what i would call fatiguing. This is still WELL within the power envelope of the Devialet (we are proabably still talking peaks of 20db).
Finally, I am reminded of a video I saw with Harbeth Monitor 40.2 and big CH precision monos. What was striking was that the monos were showing powers consistently in the several hundred watts (up to like 800 W!) range in a room that was not huge and people were standing around acting like that was not so bothersome! To me this was a clear indication that they were at the point where the Harbeths, not a sensitive speaker, were simply not getting any louder. How else could it be because it should be deafening at those levels. The speakers were clearly very power hungry to get any real sound out. How this can sound dynamic, when clearly a lot of compression was happening, is beyond me.
I don't see the point of that article with measurements given that people say tubes flatout sound louder against WAY more powerful amps, not just slightly more powerful. It's not a surprise that a high voltage device can sink more voltage into something, as they're inherently current limited for power output and not the voltage. Also the speaker driver with a tube amp can generate a surge voltage during a big impedance jump. We're seeing voltage, but that doesn't mean it's music. There's a number of things going on... Peter should have enlisted someone that understands amp design.
Meh, that depends how you look at it. Linear distortions, AKA memory distortion, is basically entirely ignored in the industry. There's no good sense to just group it in, when it's uniquely different. That's just making it even less discussed. You could have any level of linear distortion, say 10,000%, and still have 0% THD. It really has nothing to do with peaks, either. At least it isn't explicit to them, and they're unimportant to consider since they don't really change a notable amount relative to their size. All the information besides the peaks is where the memory distortion affects what we hear. And it may make peaks sound less, well, peakish.
Last edited:
I don't get your point, a reduction in the output of a driver during an amplitude peak in the music, which is comprised of an endless array of frequencies as well, is a kind of distortion...especially if it is unintentional (clearly compression can be applied on purpose...is all the time). If the music signal is 20V and the speaker produces the equivalent of only 15V of output then that reduction is compressing the peak. Now, in a multi-way speaker that compression factor will be frequecy dependent (i.e . the tweeter or woofer might compress more or less than the other driver) in all but special cases where the manufacturer was careful to dynamically match the drivers (is anyone doing this??). So, you migh getting a blunting in the mid/highs or the mid/lows, shifts in bass tuning and/or shifts in crossover as a result from a sudden burst heating up the system and dropping the output.
I don't understand where you get your example from, or where it relates to anything I've typed.
Memory distortion isn't about peaks. A peak being cut from 20v to 15v won't be perceivable unless it's a hard clip or going into a 1-2ohm load (MAYBE) - given that it's only a couple DB at most impedances. But memory distortion that comprises the general music will define how that peak is perceived. Memory distortion exists regardless of the volume you've set that may or may not introduce clipping.
We don't call clipping as distortion, because you can't distort something that doesn't exist. So it's clipped off, it's clipping, not distorted music of something that no longer exists.
Memory distortion isn't about peaks. A peak being cut from 20v to 15v won't be perceivable unless it's a hard clip or going into a 1-2ohm load (MAYBE) - given that it's only a couple DB at most impedances. But memory distortion that comprises the general music will define how that peak is perceived. Memory distortion exists regardless of the volume you've set that may or may not introduce clipping.
We don't call clipping as distortion, because you can't distort something that doesn't exist. So it's clipped off, it's clipping, not distorted music of something that no longer exists.
Is memory distortion your term? Never heard a linear distortion called memory distortion before. It will be audible if the result is a peak scaling a few db lower than one that is uncompressed...not talking about clipping. Perhaps using voltage was incorrect...I am meaning a drop in db so perhaps less current flowing and more heating as the voice coil gets hotter and hotter, which could be very rapid on a large signal peak.
Hello Moricab I think this is the article you may have seen. It is a comparison of drivers scaled in time to show how much power compression the various drivers had Worst case was 7 db As you can see it takes several seconds too loose 1 db with the best and about 1 second for the worse. This is with a 300 watt input. Granted that these are all Pro drivers As far as I know this is the only document I have seen time vs power to see the actual power compression onset and eventual stabilization 
Last edited:
I did not coin memory distortion.
We hear relativeness, so as the music perceived volume within notes and and such changes, we hear it. For example, hearing vibrato. We can perceive changes in the music this way, even though we couldn't really say how loud it actually is as the DB flys. And strong peaks will always be relatively noticeable if memory distortion is low so we can read the change happening. What makes the least difference to us is if the peak really high or sorta high, because as long as we can perceive it happening then we think it's loud unless we concentrate real hard to notice maybe it didn't truly split our ear drums. It's not a fake dynamic, I mean.
Think of it like this, you need to be able to hear the ramp-up of a sound relative to a quiet part. If the memory distortion is bad, our brains perceive the noise around the peak to be fairly high as it's a blur, and we can't track the increase very well. Because of stuff like that we get mastering of something like Rumors where one kick drum smack is like 50x louder than the rest of the album, and it sounds weird... but it's meant to impress and necessary to impress on stereos without very good memory distortion performance.
I see you want to keep the debate purely subjective, mainly on preference and your particular subjective finds of a very small numbers of cases. OK, an interesting view, but then IMHO you can not claim of physics evidence in your opinions.
My reference to formats was only to show that many people prefer an objectively compressed format to a non compressed format, not to debate recordings. But why do you say that "they took extra care in the LP production to minimize distortions" in the Valkyries" conducted by Solti? If this was just the main objective shouldn't a HiRez version sound better than the LP?
I have listened to so many exceptions to my own mental high-end rules and prejudices that I always try to be anecdotal and reporting facts as I perceive them. But when I have objective data I trust I use it as a complement of my opinions.
And yes, several Dynaudio kits sounded much better than the 1.8. I was a good friend of the local distributor in the 90's and we even built a pair in Corian, copying the Wilson's, but at that time we were not aware of his technique of controlling Corian resonance with lead!
Memory distortion was a term used once to name a type of thermal distortion in semiconductors. Gerard Perrot, also known as "Hephaistos" wrote a few articles about it in the L'Audiophile. Tubes are intrinsically free from it and a French brand, Lavardin, still manufactures SS equipment using semiconductors operated without thermal distortion.
Thanks Rob.
300W pink noise power, probably RMS (?) . How can this be relevant to audiophiles?
Besides the time scale is not adequate to discuss transients.
Yes, that is all true. Well, except there probably is some in Lavardin's stuff, just not much. One could say memory distortion for SS gear and linear distortion for transformer output tube gear, I suppose.
Hello micro
Yes that's 300 watts rms of band limited pink noise.Entire paper
http://www.cieri.net/Documenti/JBL/Technical Notes/JBL Technical Note - Vol.1, No.18.pdf
It's a real measurement of thermal compression which I find is very informative and actually helped me make driver decisions when I was assembling my active basement system. I didn't need to upgrade to vented gap cooled drivers in my application. I could live with the conventional SFG magnet structure as shown in the JBL 2225 and still achieve my SPL goal with acceptable levels of TC.
You do realize you are looking at a transient event, a specially designed one of long duration, that shows onset and stabilization. It is clearly obvious from the data that testing even the worst driver Tad 1601A with short 100 millisecond bursts of the same 300 watts won't tell you much as it may never even get to 1db. I took a full second to show a 1db drop which will be inaudible anyway.
It's only after a significant amount of time, seconds, that any of the drivers show a normally clearly audible drop in level of 3db. On peaks I doubt 3db would even be noticed. One of the better ones JBL 2241 took a full 10 seconds to drop 1 db again inaudible.
I can't think of any program material that a audiophile would subject their system to that would be anyway near as punishing.
Another reference I almost forgot about. Which basically shows the temperature rise may not be a bad as expected using program material some of us would actually listen too.
https://www.stereophile.com/content...r-voice-coil-temperatures-making-measurements
Rob
Last edited:
Hello micro
Yes that's 300 watts rms of band limited pink noise.Entire paper
http://www.cieri.net/Documenti/JBL/Technical Notes/JBL Technical Note - Vol.1, No.18.pdf
It's a real measurement of thermal compression which I find is very informative and actually helped me make driver decisions when I was assembling my active basement system. I didn't need to upgrade to vented gap cooled drivers in my application. I could live with the conventional SFG magnet structure as shown in the JBL 2225 and still achieve my SPL goal with acceptable levels of TC.
You do realize you are looking at a transient event, a specially designed one of long duration, that shows onset and stabilization. It is clearly obvious from the data that testing even the worst driver Tad 1601A with short 100 millisecond bursts of the same 300 watts won't tell you much as it may never even get to 1db. I took a full second to show a 1db drop which will be inaudible anyway.
It's only after a significant amount of time, seconds, that any of the drivers show a normally clearly audible drop in level of 3db. On peaks I doubt 3db would even be noticed. One of the better ones JBL 2241 took a full 10 seconds to drop 1 db again inaudible.
I can't think of any program material that a audiophile would subject their system to that would be anyway near as punishing.
Another reference I almost forgot about. Which basically shows the temperature rise may not be a bad as expected using program material some of us would actually listen too.
https://www.stereophile.com/content...r-voice-coil-temperatures-making-measurements
Rob
Thanks. I am not worried at all if the speaker level is 119.3 dB and not the promised 121.8 dB.
Thermal transient behavior can be easily measured using digital techniques, but as far as I know no one published data on this subject. All I know about are transient simulations, that are only validated using steady conditions.
Did you define your target SPL for music or HT? Can we know the value?
Similar threads
