Rerurn to Romy the Cat's Site

In the Forum: Horn-Loaded Speakers
In the Thread: Horns and digital crossovers.
Post Subject: Horns and digital crossovers.Posted by Romy the Cat on: 3/20/2005

This subject of digital crossovers became quite popular during last few years and many audio-people give up and go for a simplicity and painlessness of digital crossovering. There is an army of people out there who would swear by a complete transparency and non-intrusiveness of d-crossovering. Probably nowhere the subject of d-crossovering is as important as in the horn world. D-crossovering primary is so popular in the horn universe not because the crossovering itself but because the necessity to deal with delays while you do horns and there is no simpler way to introduce delays (in practically at HF) then using a digital domain. So, if people go for the D-delays then it would be very reasonable to use d-crossovers right there.

I always was a vocal and persistent opponent of D-crossovering and D-delaying. However it is virtually imposable to talk with horn people about the subject and there are many reasons why. For whatever reasons each and single audio-person is so preoccupied with the specific result that they got in thier rooms that to ask an audio-person to look at the problem wider and reflective is virtually imposable. In the end of all arguments, evidences and opinions a person usually say’s something like this say: “Come on, Romy, when I put my Behringer DCX2496 into my system and replaced my passive speaker-level, passive line-level, active-passive, active-feedback, active-subtraction and so on …. then I got a way better results”. Well, I am glad you did but:

1) There are no references how good your analog filters were implemented before – whatever I’ve seen so far was bad up to the point of being atrocious. Put in this way: there are no first-rate commercial active crossovers – no mater how expansive they would be.
2) 99% people who are advocating D-crossovers and D-delays campaigned this view using very appropriately, or at least non-seriously bult playback systems. (PA level amplification, grossly compromised horns installations and so on…) and very mistaken/primitive evaluation methods to gauge the result.
3) The domination majority of D-crossovers and D-delays supporters are not qualified properly evaluate a playback result. To bult up system using the objectives like: “When I played my jazz the my legs were dancing” or “Look, the Patricia Barbers sounded on my system so good than that my cactuses begun to blossom” is not the level where my believes operate. I know that I would endure an accusation of being arrogantly-foolish but I insist that absolutely dominating majorities of people in audio are completely dead regarding the understanding of what the deal in audio and what they do in audio.  Therefore the subjective assessments of the D-crossovering from a prospective of a foolishly-exuberant and essentially-dead audiophilic community is completely irrelevant to me. There are extremely small number of individuals, perhaps a couple hundred people around the world, who might intelligently, sensibly, judiciously, knowledgably and objectively assess a performance of playback but the rest is an ballast sound-evaluation-wise
4) Despite all bold promises from many individuals “Romy, my playback sound fantastic with D-crossovers” I NEVER experienced any interesting sound when I actually heard those systems. Nope, the D-crossovers were not a sourse of the problems in there but their entire playbacks had so much OTHER PROBLEMS and badly implemented compromises that there was no interest to even think about cons or pros of D-crossovering in context of those inhalations.

I have to admit that I would LOVE to see a digital solution to work well and if I find that it is the case then I would be gratefully embracing it. Unfortunately it never happened so far. My experience with this started a couple years ago when I discover an absolutely phenomenal mid-bass driver to bult 40-50H mid-bass horn. However, I needed a low-pass right after it does 1-2 octaves and to introduce 8.5 feet delay. Certainly I would not use a high-order analog filter and to introduce the delay for wide-bandwidth HF would be a sonically suicidal. So, I began to explore the D-solutions.

Whatever I tried did not work and I am not talking about crossovering but juts AD-DA conversion: an introduction of ADC-DAC element into a signal path instantaneously screwed up sound. The very minute deviation between the shadows of the tomes became less distinctive, the inflection of voices and pitches became less characteristic, the musical accents and enunciations became muted and more neutralized, music become more genetic and less idiosyncratic, more washed-out. In other words the result defeat the purposed that I was pursuing with my audio objectives. Put in this way if I conduct an orchestra that would play this way then I would fire the musicians, so why should I go in there with my playback? All that I was describing happened with all 5-6 processors that I trued. All of this affectively removed among the available options the idea of D-delaying. The only one solution that I fine interesting, was to have a pair of DACs after my CD transport connected in parallel and to bult-in a D-delay into one of them (his would not introduce any problems). I would certainly go there but I do have an analog setup that I play quite aggressively. What should I do with analog? So, I ended up with a fiasco to find a good solution for delay in reference to my mid-bass horn and the horn building ceremony never was initiated. The “phenomenal upper bass driver” is sitting in my storage unit the 21-century civilization find a way to make delays that work :-) or until I will be able to do it naturally by appropriate positioning of the horn

Now the funny part: the D-crossovers. As that time I got the Behringer DCX2496 and decided to experimented with it crossovering possibility. As soon I activated the crossovering the result was even more catastrophic then I described in previous paragraph. It was like someone inject fog into music, bleached it out, compress it, along with countless other problems. All those problems were of the magnitude that there is no needs to talk about anything further but juts trash the crossover and never touch it. Later on, I borrowed form my local pro shop all-possible the most expansive crossovers and the result was identical: as soon the curve was introduce the sound when down. The very same was experiencing when I played with countless CD recorders – as soon I touched that fade option or anything similar DSP-based, then sound went immediately into a toilet. My experience with very many preamps that had digital remote controls was pretty much identical, so I decided that it was not the direction to go.

Then I begin to approach a number of the individuals, the world-class digital designers who designed whatever you today considered the best in the digital domain. They expanded me (and this was a very extraordinary occasion when ALL OF THEM were agree on this subject UNANIMOUSLY) that DSP, no mater how good it implement, even for with no price restrictions, is a fundamentally sound screwing concept. As soon a curve introduced it done by removing bytes at the bottom of the curve and as a result it introduce problems of loosing resolution and dynamics at lower levels. Country to what thy explained me I clearly hears a severe problem not only near the slopes but even at band-pass. Nevertheless, they all (all of them!!!) told me:  “It is not about to bult a properly sounding D-crossover but to make it that a given user with his/her reference points would not be able to become aware of a sonic degradation”. I would not bring the names of those people to back up my story because some of them today … are manufacturing digital crossovers and some of them are about to release D-crossovered loudspeakers. However, I might assure you that they are the most respected and the most competent names in today digital sound possessing. I have to note that I was very serious about the project and I was ready to pay $20K-$30K for a digital delay-crossover that would not introduce any noticeable by me deterioration into a signal path. When the designers learned about my objectives they suggested me that I would not find a solution with their digital world. So, I abandoned this direction.

Now, some further observations on the subject and why I consider that generally people who dream about the beauty of DSP processing with horns looks at fundamentally wrong direction and think kind of wrongly.

First off all: why do we, the horn people, ever need a digital crossover? (I am not talking about the delays now). People who know me know that I religiously multi-channel guy and that I believe that so-called full-rage channels is a celebration of a wishful thinking and sonic idiocy over really and actual results. So, the only know to me way to make a playback sound acceptable in context of multi-channel is to design the channels that have no-electrical low-pass filters (use acoustical filters, throat EQ or a nature driver decay) and use no higher then first-order electrical high-passes. Let me to repeat: 6dB per octave eclectically – nothing else!!! So, we have one single cap that should be used in the entire channel path. So use it! Place it wherever you wish, at speaker level works very fine, way-way-way better then any imaginable D-crossover. I have to report the even at speaker level a capacitors, although it might have some different “sounds” and tonal influences, but the NEVER destroys the IMPORTANT attributes of sound responsible for musicality. The speaker level capacitor affects audiophilism but it dose not screw up music.  If you wish to go further then instead of a speaker level cap place instead of the across-stages coupling cup if your amp use them and the design permits. Alternately you might run this single serial cap against a high-impedance of your line level stages, consequentially reducing it’s value to pF or nF level and using some high-quality caps (air, teflon, vacuum and so in). The little tonal deviations that they might have are completely irrelevant at speaker level and of course it you place a superb quality ultra-low values cap at line-level then you have absolutely no-deteriorating result. So, why in the hell we need a digital crossover?

Some people would disagree with me. Those people, primary the GOTO, ALE, TOA and BMS users, would suggest that a proper horn-loaded reproduction is imposable within a channel operating with a none-limited range and therefore, by used a multiple very narrow bandwidth channels, we have to use a higher-order crossovers; so, considering that the 2-4 order crossovers are imposable to use at analog domain we have to go digital.  I would very much argue this point of view. The RCA’s Radiotron clearly suggests in 20.4 that limitation of frequency ranges relates to a desirable efficiency and my experiments do confirm this observation. With a park of existing drivers is it perfectly possible to accomplish 110dB of sensitively via 4 channels. As soon you go higher then approximately 110dB-11.5dB then you do need to introduce one more channel but do we REALLY need to go there? The GOTO and ALE would be happy to sell you 10 differents drivers for each single octave each, ironically archiving no more then 110dB sensitively but forcing the system users to buy more drivers (it never hurts, doesn’t it), facing the person to face a crossovering nightmare and the dilemmas of the horns alignment/positioning. So, do the artificial believe in a horn systems with 453 channels has any practical rational? I do not think so. Therefore, I do not see any arguments for building 5-6 channels systems. It would be ironic to mention that the word suppliers of GOTO and ALE drivers are… “accidentally” the distributors of Behringer and BSS digital crossovers.

(I would accept, and agree with some arguments about the excessive amount of LF channels: like low-LF, mid-LF and upper-LF channels but with the LF is totally different story and here is where the higher order crossovers might be utilized. However, bass channels is totally different conversation)

So, were I am staying? Certainly I propose that the digital crossovers should not have a place in horn installation if the system is designed rationally, knowledgeably and with a respect to the Result. They were my words, my experience and my thinking. Your mileage might and most probably will wary. However, so far I did not hear any result that would encourage me to review my position and all installations that I heard (not only horns) that used D-crossovering were at sub realy serious level.

Romy the Cat

PS: I have to make this disclaimer because most of Stereophiles are sick on their heads and would consider this article as a personal attack on their damn playbacks. The Morons: I hope you understand that this writing has a pure conceptual and educational purpose and it does not call you to take a hammer and crash your d-crossover. Do, think and believe in whatever makes you happy and if you do used a digital crossover with horns then it is YOU who benefited with the result. Perhaps you "deserve" it….

PSS: For whoever who does use the D-crossovers:

Sometimes ago I argued the point with a French guy: Jean-Michel Le Cleac'h somewhere on the Web. Unfortunately that conversation went to nowhere because that site where the conversation took place was moderated by some kind of idiot who in order to kiss that French guy's ass was deleting the posts that disagree with Jean-Michel’s view. Anyhow, if you do use the Behringer DCX2496 or BSS FDS388 them you most likely experiences the very crappy sound from your D-unit (it is what I always heard from YOUR systems). Jean-Michel offered an explanation for it. I did not try it (when I used the D-filers I was resolving the unity-gain via different means) but his explanation sound quite reasonable and you might consider it.

Here are the Jean-Michel comments:

“When I first replaced my previous active Kaneda crossover, which one used discrete stages and was excellent with the BSS FDS388 I was really disappointed. The sound was dirty, noisy, harsh... then I measured how many volts RMS I had on the input: 80mV. This meant that I had +37dB digital noise added. Then I put step down transformers and attenuators on every outputs and feed the inputs at high level (attention to clipping!) this was the best crossover I ever possessed. Please notice that I have also transformers at the input and that the crossover input lines and output lines are balanced.”

“A device like a digital crossover when used with analog inputs and analog outputs possess a gain of 1 (well in fact there is also a digital attenuator included generally but if you use it you loose also some bits and some precision in the output signal) I don't know what is the sensitivity of your amplifier but most surely for the max acoustic level in your listening room the input signal of your amplifier is probably less than 1V RMS. This means that you have also 1 volt RMS at the inputs. To obtain the best accuracy of a digital crossover as the Behringer DCX2496 you have to feed it at the inputs with a max signal of 10 VRMS. If you feed it with a 1 V RMS max you loose 20dB on the ADC and the DAC. Our experience is that this bad use of the ADC and DAC overall dynamic is the main explanation of the bad sound that few reported with digital crossover. To use a digital crossover you have to use the correct input level and this often means the use of a larger preamplifier gain. In consequence you have also to put an attenuator at the output in order that the power amplifier input doesn't saturate.”

Rerurn to Romy the Cat's Site