GoodSoundClub - Romy the Cat's Audio Site

Horn-Loaded Speakers
Topic: Just 3 questions.

Page 1 of 1 (17 items)

Posted by Romy the Cat on 08-03-2013

TwoGoodEars posted at his blog a report about is listening session or some kind of installation. I still have no idea who’s installation it is: Jean Hiraga or Be Yamamura. Nevertheless, I think it very much worth to look into it.

http://twogoodears.blogspot.de/2013/07/jean-hiraga-meets-ale-and-be-yamamura.html

I presume if Jean Hiraga is involved that it is all 12 channels of Hiraga SS amplification. It is 6 ways with bass drivers but I can see only one upper bass horn witch is in my estimation is about 60Hz-70Hz. Where is the mouth of lower bass horn and how low it goes I do not know. The owner looks like stick to all ALE drivers and Presume he knows them well. No valuable data is provided by TwoGoodEars – no idea how he mitigated the Fs of the bass drivers for the specific horns.

The upper horns are straight forward and made from metal with “exotic and elegant oxide”. I am not a big fan of metal horns and the mentioning the “exotic and elegant oxide” sound like BS to me.

The horns are stuck nicely in the corners of the given room and the long extension of the bass horn looks like smartly organized. The horizontal offset of the channels axes of course is a big no-no in my book but this setup has a secretive tool to fight with it - 20th order (120 db/octave) crossover. With this croosovering any crosstalk between the channels does not exist. How the horizontally offset channels are integrated with each other in this configuration? I have littlest ia in this case as I never can imagine how 120 db/octave would sound. Obviously this brick wall filter has very violent phasing problem but it is line level against fix impidance and they might be able to hold it somehow by attempting to run phase-minimal slopes. I do not know and TwoGoodEars of cause is silent about it.

The use of digital croosovering of cause is the most questionable in context of this system, particularly in the case of ALE driver that have high capacity t react to minimum impedance. With 120 db/octave slopes even if they use 24 bit depth they will have at minus 6dB some kind 8 bit resolution, almost like a telephone sound. The question is: will be able to hear and to recognize that minus 6dB in context of 120 db/octave slope? I do not know the answer.

Generally I very comfortably and accurate can predict some characteristics of playback sound just looking at the pictures but in case of this system I do not know. For sure it is close-bottom system and it will not have “that” bass but that is not something that I would consider important. IT would be interesting to hear this installation.

The Cat

Posted by twogoodears on 08-04-2013

Hi Roman... the installation is a combined joint-venture between Endo-san of ALE (bespoke drivers and horns) and Be Yamamura's (45 W solid state Class A amps and digital software and hardware).
Jean Hiraga, like myself, were only invited to taste this system and the ongoing fine-tuning, yet he wasn't involved in ANY way in the system making.
The listening session was both satisfying and (sort-of) limited, as the digital library was in the making and the tuning of the system still was a work-in-progress, due to minimal adjustments mostly performed in drivers/horns delays and integration.
I've been told by Be Yamamura that phase and delays between the ways are managed via the powerful software (and usual measuring mike, etc).
The interesting question about digital crossovering will be asked by yours truly directly to Yamamura-san, soon: more details about frequencies cut-offs, drivers used will be implemented in the Blog asap, as given by the system landlord himself.
Sound... mmmhhh... it's natural, effortless, not so deep bass, meaning the sense of stomach punch associated with woofers isn't here... a light, airy low end is here, instead.
BTW: the system owner told me about his next step - i.e. adding another tweeter, bringing the system to an awesome 7 ways, with very specialized drivers and ultra linear and limited bandwidth... I expressed some doubt about redundancy and loosing the target risks.
Anyway, it's a system worth a listen, Roman... workmanship is top class, cost-no-object and I'm sure the result when such a complex system will be tuned will pay-off the (huge) investment.

Posted by Romy the Cat on 08-04-2013

Thanks, twogoodears. I do not particular care about the frequencies cut-offs as I do not know the ALE drivers. If I speak with the system owner in context of what you say then I would post 3 questions:

1)    What specific limitation or shortcoming of current sound he would like to overcame by introduction of another tweeter
2)    What specifies sonic, design, configuration, integration or any other advantages the own recognize by going with 120dB per octave. What would be different if the system use let say 24dB per octave?
3)    Does the owner srecognize any negative contribution of digital crossovers?

The Cat

Posted by haralanov on 08-05-2013
fiogf49gjkf0d Hi Stefano, why don't you kindly invite Mr. Be Yamamura to join this forum? Anyway, I want to add another simple question to the list: Did he tried another channel's arrangement (for example tweeter positioned below midrange channel or some other configurations)?

Posted by Romy the Cat on 08-05-2013

The more I think about it the more I realized that this 120dB per octave crossover is very interesting move. For sure it is bold move. There are unquestionable benefits of such a “barbaric” crossover as the throwing way bits at the signal get attenuated has virtual no room to be heard. Also, there is no lobbing of multiple sources of any kind, neither the comb filtration distortions. So, hypothetically if you have 6 channels, run only digital and have 6x120dB per octave crossovers build into 6 identical DACs then you do cover all grounds and it is possible that in THIS application digital crossovering would not be as bad as it uselessly is. Still, I have absolutely no idea how two drivers located at tow different axes and separated by 120dB per octave crossovers would sound. I in my practice so much accustom to recognize an individual channel is not an individual channel but rather a combination of this channel with two neighboring channel that I am a bit at lost to forecast how 120dB per octave channel would sound. I do know that this 90 degrees slope does not exist in nature and most likely crate some kind of phase big dang but I would not guess anything else.

Posted by Romy the Cat on 08-06-2013

I still thinking about this ides of ultra sharp digital filtrations and for whatever reasons it fascinates me. It is not the I am consider to do anything with it – it is absolutely not needed for me – still I might like the idea as a topological solution. Some regular digital crossovers I think do allow 20^th order. If you remember in the London’s thread a few days back I told that I would like to hear some French installations. There is, or at least there was, a guy in France, his name is Jean-Yves Kerbrat. He has a large room and he piles a LOT of exotic Japanese drivers in Macondo Configuration and all digital crossovered. I of cause discarded this practice but I never thought though that it might be 120dB per octave that might (or might not) change the rule of the game. I need to call to my digital experts and ask them what and how the120dB per octave filter is implemented on digital domain. I think TACT has 120dB per octave and it would be fun to hear it. I still feel that the “digital” might be heard even with such sharp filter as I did participate in the experiments where I was accurately pointed out the files where dither was applied (at -132dB!!!) . Still, how the driver will sound with 20^th order is VERY interesting even that I do not know about the filter ringing consequences with such a step slop …

Posted by zztop7 on 08-07-2013
fiogf49gjkf0d Just replying so I get email updates.Very interesting.zz.

Posted by twogoodears on 08-08-2013

Yes, maybe the 120 db/octave is THE trick for complex multi-ways systems of highest resolution, yet this barbaric:-) slope can ONLY be achieved by digital domain crossovering... unfortunately, as a satisfied user of TacT RCS 2.2 XP AAA room-corrector/crossover/DAC/preamplifier, the up-to 24th order crossovering can only be operated in sub/low-end area - i.e. any frequency to blend bass and mid-low, but ONLY this.
That's - among others - the greatest strength of new Be Yamamura's DAC/crossover/music server/software... Lyngdorf is still more limited, only average automatic room correction capabilities and no crossover whatsoever... and, then there is Groundsound, well... maybe it's worth digging it, as the latter allows better digital crossovering than TacT's, but MUCH limited hz-by-hz room correction and time-delays vs. TacT's, still the best... 'til Yamamura's appearing.

Posted by speedysteve on 08-08-2013

The systems I've heard with steep slopes (only 4th or 8th order) sound hard and unnatural to me. Admittedly these are on everymans drivers - the sort of thing we all can buy and deploy.Several digital X/O users I've talked too all seem to gravitate to 4th order for all. I've tried this on my system and it just does not sound as good to me as selected slopes sympathetic to the driver and and horn application and room too.I have found 4th order to be the order of the day, for the deep bass channel. 2nd order for mid bass and 1st order for upper side of the mid. 1st order for upper mid and tweeter. The latter two with no upper slope at all, just playing open - less is more.I moved from 2nd order on the upper mid to 1st order on getting the S2's and experimenting with them and reading Romy's own very informative texts on the S2 and 3uF cap listening tests and measurements.I don't recall the exact quote but something like - a really good driver is doing things far down the SPL scale that a poor one could only dream of...
I'd love to hear 120dB/oct system just to know what it sounds like - with the right implementation and due care to everything I could envisage it working from a physics pov, but in the real world?

Posted by Romy the Cat on 08-08-2013

twogoodears wrote:
Yes, maybe the 120 db/octave is THE trick for complex multi-ways systems of highest resolution, yet this barbaric:-) slope can ONLY be achieved by digital domain crossovering... unfortunately, as a satisfied user of TacT RCS 2.2 XP AAA room-corrector/crossover/DAC/preamplifier, the up-to 24th order crossovering can only be operated in sub/low-end area - i.e. any frequency to blend bass and mid-low, but ONLY this.
That's - among others - the greatest strength of new Be Yamamura's DAC/crossover/music server/software... Lyngdorf is still more limited, only average automatic room correction capabilities and no crossover whatsoever... and, then there is Groundsound, well... maybe it's worth digging it, as the latter allows better digital crossovering than TacT's, but MUCH limited hz-by-hz room correction and time-delays vs. TacT's, still the best... 'til Yamamura's appearing.

Twogoodear, this is not about the devises that do filtering or about the mythical “software” that does post-filtering correction. This is rather my attempt to get a feeling of the new topological move, something that I did not see yet before and something that I have no experience to hear before.
I use the word “barbaric” very deliberately as this is in a way very unnatural way to implement crossovering. This bold way do take care about the worsening of sound that uselessly come from loosing the bits during the slope and this is very good. However it shall also bring a whole arrays of own problems. I know it will but I have no idea how those problems will manifest itself auditably in context of playback.

Those brick-wall filters shall ring like crazy and 120dB per octave shall ring insanely. How will it be auditable I do not know as the brick-wall ringing in context of that system is spread across the channels. With absolutely unpredictable phase shit of each filter I do not know how they manage the system.

Now the most interesting part. If all 6 crossovers (12 slopes) are managed by one possessor then theoretically it is possible to analyze the ringing and phase discrepancies of the each individual slope and compensate it on the contra-slope. People use it in HF brick-wall filtration but the load in this case is a cable or passive and “flat” load. In our case the “load” is 12 compression drivers dispersed in space and analyzer is a human brain that is good 10 feet apart. So, even theoretically it is possible to equate the whole 12 channels delays, phases and ringing patters only for a single point in space. I am not taking about the sweat-spot about wish we accustom to talk in audio. I am waking about a few mm deviations from calculated sweat-spot and then a very sharp 120dB volatile chaos that come outside of that sweat-spot. It is mathematically impossible to balance the channels for 2 ears it might be only for one. Furthermore it is imposable to do it for continue octave but only for one differentional frequency. So it an instilment has let say a continue glissando then the drivers shall have a distinctive slide from in and out if phase cancelation.

This is all theory and I do not know how it sounds in practice. I would like to hear it and to confirm or to refuse my guess. What county the playback is set up?

Posted by Romy the Cat on 08-08-2013

speedysteve wrote:
The systems I've heard with steep slopes (only 4th or 8th order) sound hard and unnatural to me.

Yes, the higher order crossovers do sound harder even if all time alignment measures are taken. Still, do not forget that time alignment measures with horns and high order crossovers are kind of tricky. We decay a channel and we understand that sometime the decay of the driver kick in at the bottom of the crossover. What we frequently forgets that the decay of the horn comes before it and the horn decays with good 2-4 order. So we have let say a MF driver with 1.2K second order crossover point sitting in 500Hz horn. Just a bit over octave from the crossover point the horn begin own 3 order slop, then the driver kick in with good 4-6 order. So, in phase domains we are very much screwed with horns. So, what I feel is important is not only the more or less consistency of the low orders complimentary slopes but also the proximity of the sloped to the drivers natural boundary. You do not want to minimize the situation when one driver operated at it’s bandwidth capacity and another complimentary driver is set at super comfortable 2-3 octaves from own natural roll off. Sure you not always can accommodate it but it need to taken under consideration.

Posted by twogoodears on 08-09-2013
fiogf49gjkf0d ... indeed, your approach is illuministic and curious... it would great you'd taste this system with your own ears and sensibility and listening biases (unbiased...)...The ALE/Yamamura's system is in Bologna, Northern Italy and a listening can be arranged... PM, etc. etc.BTW, the uncontinue octave cut-offs also made me asking about musicality and naturalness... yet...

Posted by cv on 08-09-2013

I bet Yamamura is using non-causal filtering... this is all running off a music server, so could process everything offline.

This makes it possible to do forward-backward filtering - you run the filtered signal "backwards" through the original filter (ie reverse the FIR coefficients) and get zero phase shift....and doubled up amplitude response...

Scale everything appropriately to make best use of 24bit dac range (easily done offline), add some proper dithering, and Romy's objections to digital filtering might well disappear...

Posted by Romy the Cat on 08-09-2013

cv wrote:
I bet Yamamura is using non-causal filtering... this is all running off a music server, so could process everything offline.

Chris, there is no such a thing as “non-causal filtering”. Filtering is filtering, would it digital or analog, on digital domain of cause it is much easier to do 20th order. The fact that Yamamura uses music server does not say anything. He might have the filtration splitting before DAC (the way how it shall be) but then he needs to have 6 DACs. From the Stefano original article on his blog I understand that he does not use 6 DACS, so it is most likely configured like this: DAW-> I/O interface - > DAC -> crossover (with own D/A and A/D) -> amps. It might be different configuration and he might use own software based filtering of the DAW’s stream but then how the software interacts with analog power amps?

cv wrote:
This makes it possible to do forward-backward filtering - you run the filtered signal "backwards" through the original filter (ie reverse the FIR coefficients) and get zero phase shift....and doubled up amplitude response... Scale everything appropriately to make best use of 24bit dac range (easily done offline), add some proper dithering, and Romy's objections to digital filtering might well disappear...

Well, this is exactly the propose of this mental exercise – to assess if it possible to have the Romy's objections to digital filtering disappeared. I think when you think about what Mr. Yamamura you mentally employ forward-backward digital filtering-correction. For sure it might be done but there are two major conflicts in this direction:

1) The ringing the take please with brick-wall filtration is too short to be handled by conventional digital processors. Any processors run at some kind of sampling rate, let pretend that it is 192kHz PCM. Let do not look into the DSD or Double-Rate DSD as it is all crap as they do PCM editing at 352,800 Hz. If you look at the scoop then you will see the ringing at brick-wall filtration is too fact to be processed even at 192kHz. The ringing distortion are just sit between two individuals samples and therefore then can’t be corrected.

2) This is the biggest problem of all. What you think about backward reconstruction through the original filter then think what would be the reference of reconstruction. For sure it would be a complimentary slope that has to be matched, right? Well, the Mr. Yamamura configuration has no such a slope. If you split a digital steam on 6 channels, crossover them and then combine them together into one cable then you can do analyses of the complimentary slopes, recognize what does not match and do the corrections (still within the limitation of the current sampling rate). However, in case of the inhalation above there is no cable into which the 6 signals shall flow but instead it is 6 individual drivers, dispersed in space and surrounded by air. All of it has distortion rate way less comparable with what backward reconstruction might understand.

OK, let to fantasize a bit. Let pretend that Mr. Yamamura has a microphone at the listening position that has 6x120dB per octave partitions and do analyses belong the same bandwidth as the original crossover do. Let pretend that during calibration phase they run squares across the system and the microphone “get” the ARRIVED discrepancies. We can presume that manipulating the crossover and delay we could tweak something to make the arrived pulses to look more acceptable. That would be cool solution, it is possible to do and… this is exactly what I would like to hear as I do not know if it has any practical sonic merit. It needs to be heard…

Rgs, the Cat

Posted by decoud on 08-09-2013
fiogf49gjkf0d Zero phase filtering would not require a reference: the signal is just run through the filter in forward and reverse. But there is another filter to consider here, a physiological filter - presbyacusis - that means Yamamura's hearing threshold above 10k Hz is probably 60dB or more. So maybe he does not care about the ringing....

Posted by Romy the Cat on 08-09-2013

I do not believe in applied presbyacusis. It is not that I do understand the subject of loosing spectral sensitively with age but I certainly do not feel that deterioration of HF sensitivity makes a person more or less capable to register music. We in audio invented to ourselves a myth that HF sensitivity matters too mach but it is not.

Also, the ringing that uselessly accustoms very sharp filters does not manifest itself as HF noise that might or not be heard by older people but rather as general hardening of sound, creation of different echoes and particularly prior to the transients, destruction of imaging, softening of lower bass.

The caT

Posted by decoud on 08-09-2013

It is a more complex filter than that: the change in HF is just what we measure to detect it. Clinically, one never looks at transients, for example, though it is likely that the fundamental mechanism affects them too, for what you lose in the broadest terms (at least the neural component) is dynamic range per unit time. It is striking, for example, that speech comprehension in conditions of noise is affected even when the measured HF attenuation is only just within the reported frequency window of speech. I don't know if anyone has studied it properly but I suspect it is precisely because you lose the capacity to deal with the fast non-stationarities that are characteristic of speech.

Page 1 of 1 (17 items)