I think many balk at the very idea of supertweeters because they are committed to an ideology such as "phase perfection" (new name I just made up).
Phase perfectionists want zero phase shift in the audible range if not far beyond. The want total pulse coherency, with step-type pulses reproduced perfectly in the time domain.
The arguments for this go back nearly a century, since it became clear that it was difficult to design audio equipment, especially loudspeakers, without significant amounts of frequency-dependent time delays, often thought of as "phase shift" (thought time delay is a more neutral description).
Those arguing against phase perfection generally claim it wouldn't be a bad idea, but it's too expensive when it comes up against other design considerations which are considered to be more important, such as dynamic range and bandwidth.
Blind testing has generally shown insensitivity to small to moderate amounts of frequency-dependant time delays, like those caused by most well designed loudspeakers. Many designers, such as Sigfried Linkwitz, explicity claim that the phase shifts introduced by crossovers such as the 24dB per octave Linkwitz-Riley are inaudible, and it may have larger time delays than simpler crossovers.
But phase perfectionists stick to their guns, claiming they hear the difference between speakers that use true first order acoustic crossovers (such as Vandersteen and Thiel) and comparable designs. Phase perfectionists use either single drivers, often planar (such as Quad ESL-63 and its successors) or dynamic systems with 6dB/octave or "crossoverless" design. (Crossoverless designs put a high pass network, often just a single capacitor, in series with the tweeter but let the woofer(s) roll off naturally.) Often these systems are not as phase perfect as their advocates believe they are. Most speakers which claim 6dB/octave crossover only achieve something like that electrically; acoustically their performance may be anything but 6dB/octave. For example, if a 6dB/octave crossover is combined with the 12dB/octave highpass response of most tweeters, an 18dB/octave crossover results, only being staggered somewhat by the tweeter having much lower highpass cutoff frequency.
If you count products in the marketplace, or satisfied listeners, or cumulative reviews, it is clear that the high order crossovers despite all their frequency-dependant time delay, are by far the most popular. Right now the 24dB per octave Linkwitz-Riley may well be one of the most popular crossovers in expensive loudspeakers ($1000 and up). Cheaper speakers economize with 12dB per octave.
Anyway, when I set up my supertweeters last year I did indeed examine pulse response carefully, if not exactly requiring an ideal "perfect pulse" I at least wanted a brief and distinct pulse-like waveform. You get some sort of pulse picture during the Tact RCS setup. I was able to make some adjustments to the supertweeter delay to make it blend in. Though I would not claim I am achieving anything like phase perfection.
Sufficiently above it's highpass point of 14kHz or thereabouts, the Elacs are in polarity with little phase shift. It is only around the crossover point where there is some phase funniness as they blend with the upper response of the Acoustats. I would not expect the Acoustats to have phase perfection up there either; they seem to have a peak around 13.8kHz.
You could say that both supertweeter advocates and phase perfectionists claim the need for things that haven't generally been proven to be needed. And although the two approaches could in principle be achieved, most often phase perfectionists will balk at the need for the kinds of compromises which would usually be required for supertweeter integration, and supertweeter advocates would find it nearly impossible to pursue their approach and phase perfectionism at the same time.
Friday, January 28, 2011
High Level (+15dB) Supertweeter level wins again
In past few posts, I've been describing the "more defined bass" effect that seems to result from using a supertweeter. I did not invent this idea, you can find it in many other blogs.
And there is another claimed effect from supertweeters. It is claimed they do not make the sound more harsh. That is the weak form of the claim. The strong form is that they actually reduce harshness.
If you look into the claims made for supertweeters, these and many more are common.
Others scratch their head with incredulity, and say that they can't hear as high as their current tweeter can play, so why do they need something that plays higher? Others say they may get too many highs, or at least too much stridency and harshness, as it is, so why would they ever want more (and apparently they don't believe the claim that supertweeters do not increase and may decrease apparent harshness).
Many of these skeptics have not even heard any supertweeter. And since most supertweeter setups are consumer addons (though I'm just guessing, it could be the other way around), or even otherwise, it can always be claimed that the supertweeter setups they have heard were not "properly integrated". Supertweeter integration is a large topic that you could read about endlessly.
On the other hand, I'm not aware of any blind tests, and it may not even be possible in many cases to do a blind test. It is likely beyond the means of most hobbyists.
Well I never say my tests are ultimate and definitive, but I continue to sense the supertweeter as decreasing harshness, though sometimes adding a distinct unwanted sonic signature or localization.
Last night I did another test of the +10dB and +15dB crossover levels for the supertweeter (at the current 15.5khz crossover setting). Listening to "Lucky Man" from Emerson Lake & Palmer, I was thinking Greg Lake sounded a bit strident. (A lot of that, no doubt, is the recording, intended to give Greg an otherworldly sound.) So I then changed to the +10dB level. No particular change in Greg Lake's vocals, but now the highs seemed more chaotic from all over the soundfield, there were lots of distractions in the high frequency range. Restoring the +15dB level restored sanity, though perhaps you could claim a kind of totalitarian simplicity, with Lake's voice commanding a swirling mix of instruments.
I have now read some claim the Elac 4pi tweeters can sound metallic if played too loud, and I am struggling with that. To get the most good effects from the supertweeter, it has to be played loud enough. Perhaps to get any good effects... But then, at times the supertweeter can give itself away. The supertweeter can be "hidden" completely by making it's level low enough. But then it doesn't seem to have the good effects. So there is a fine line between not-loud-enough and too-loud, and sometimes one can wonder if there's a line at all.
I think there is, unfortunately, a tendency toward metallic sound from the Elac. If the Elacs were perfect, like the pulsating plasma in Hill Plasmatronics, they might not have that sound. The sound can possibly be tweaked with DSP, class A amplifier, cables, etc., etc. But it complicates the integration problem.
Speaking of Plasmatronics, I had a friend who was combining the plasma driver with Acoustats back in the late 1970's. It sounded great then. But I don't consider the Plasmatronic driver to be particularly practical and I'm not even sure I'd want two in my living room.
And there is another claimed effect from supertweeters. It is claimed they do not make the sound more harsh. That is the weak form of the claim. The strong form is that they actually reduce harshness.
If you look into the claims made for supertweeters, these and many more are common.
Others scratch their head with incredulity, and say that they can't hear as high as their current tweeter can play, so why do they need something that plays higher? Others say they may get too many highs, or at least too much stridency and harshness, as it is, so why would they ever want more (and apparently they don't believe the claim that supertweeters do not increase and may decrease apparent harshness).
Many of these skeptics have not even heard any supertweeter. And since most supertweeter setups are consumer addons (though I'm just guessing, it could be the other way around), or even otherwise, it can always be claimed that the supertweeter setups they have heard were not "properly integrated". Supertweeter integration is a large topic that you could read about endlessly.
On the other hand, I'm not aware of any blind tests, and it may not even be possible in many cases to do a blind test. It is likely beyond the means of most hobbyists.
Well I never say my tests are ultimate and definitive, but I continue to sense the supertweeter as decreasing harshness, though sometimes adding a distinct unwanted sonic signature or localization.
Last night I did another test of the +10dB and +15dB crossover levels for the supertweeter (at the current 15.5khz crossover setting). Listening to "Lucky Man" from Emerson Lake & Palmer, I was thinking Greg Lake sounded a bit strident. (A lot of that, no doubt, is the recording, intended to give Greg an otherworldly sound.) So I then changed to the +10dB level. No particular change in Greg Lake's vocals, but now the highs seemed more chaotic from all over the soundfield, there were lots of distractions in the high frequency range. Restoring the +15dB level restored sanity, though perhaps you could claim a kind of totalitarian simplicity, with Lake's voice commanding a swirling mix of instruments.
I have now read some claim the Elac 4pi tweeters can sound metallic if played too loud, and I am struggling with that. To get the most good effects from the supertweeter, it has to be played loud enough. Perhaps to get any good effects... But then, at times the supertweeter can give itself away. The supertweeter can be "hidden" completely by making it's level low enough. But then it doesn't seem to have the good effects. So there is a fine line between not-loud-enough and too-loud, and sometimes one can wonder if there's a line at all.
I think there is, unfortunately, a tendency toward metallic sound from the Elac. If the Elacs were perfect, like the pulsating plasma in Hill Plasmatronics, they might not have that sound. The sound can possibly be tweaked with DSP, class A amplifier, cables, etc., etc. But it complicates the integration problem.
Speaking of Plasmatronics, I had a friend who was combining the plasma driver with Acoustats back in the late 1970's. It sounded great then. But I don't consider the Plasmatronic driver to be particularly practical and I'm not even sure I'd want two in my living room.
Thursday, January 27, 2011
Bedroom System loves DVD-Audio and HDCD's
Last night spent some time listening to Bedroom System, notably the DVD-Audio of Queen's Night at the Races. Wow! The dynamic range on this DVD-Audio presentation (I used 96kHz PCM stereo mode) is fantastic!
Last year Bedroom System got a new preowned universal disc player, a Denon 5900. This is one of the all time great universal disc (except Blu Ray) players, featuring one of the best Burr Brown DAC's, 120dB signal to noise ratio, and separate power supplies for audio and digital. Weighs almost 30 pounds, originally cost $2000.
Unfortunately, I was not as lucky with this second purchase of a 5900 as I was with the first which went into my living room. The second was not as clean looking, and has a noticeable whine on high speed discs like DVD-Audio and SACD. It may need internal cleaning and/or a new bearing or drive mechanism. Mind you, the unit works OK, it plays every disc I throw at it, it's just more mechanically noisy than a top shelf unit should be.
I figured I might be able to get the noise fixed at the Denon service center about 100 miles away at probable cost $100-$250.
But I had a cheaper solution. I simply avoided retiring my previous bedroom player, a Denon 2900. Very similar to the 5900 except lacking the dual power supplies and also lacking HDCD. It uses the early-production version of the same DAC as the 5900. This 2900 I bought a few years ago is minty clean.
So what I do is this: I play low speed discs, CD's and HDCD's, on the 5900, and I play high speed discs, DVD-Audio and SACD, on the 2900. Only the 2900 needs a video connection, and it connects to video monitor using S-Video through a isolation transformer to prevent contaminating the audio.
This is only for actual discs. Mostly I listed to computer files through my Sonos system. But Sonos doesn't handle high resolution formats, and I don't necessarily have all my discs copied to Sonos (copying discs to an online library is an unbelievable amount of work if you are somewhat perfectionistic about it).
In the long run, I have two nearly identical machines, so one could serve as spare parts for the other. I've even thought about transferring the quiet mechanism from the 2900 to the 5900, but for the time being, why bother?
Both players feed analog into my digitally processed system. For many years, I have used a TacT RCS 2.0 preamp in the bedroom, which feeds Behringer DEQ 2496 and DCX 2496. Unfortuantely, the analog input on the Tact features a somewhat outdated analog-to-digital converter which effectively only has about 16 bit performance.
But late last year I bought a new analog-to-digital converter (ADC). It's actually another Behringer DEQ 2496, which I plan to use for other purposes also, but inside the DEQ is a fairly decent AKM ADC.
I've had this hooked up, but not very well until last night. Previously I ran the Denon players through my Acurus L10 preamp to set the level, then into the DEQ. But last night I decided to bypass the gain stage and volume control of the Acurus, and use the tape output instead. That seemed to really open it up, just like the living room system when I play advanced resolution discs.
And of all the kooky things I was not connecting the Behringer "ADC" to my Tact very well. I was using the optical output of the Behringer. But then it had to share the optical input on the Tact (which has only one optical input) with my carousel DVD player. Last night I got tired of switching the optical cables, and decided to connect the Behringer using AES/EBU instead.
Last year Bedroom System got a new preowned universal disc player, a Denon 5900. This is one of the all time great universal disc (except Blu Ray) players, featuring one of the best Burr Brown DAC's, 120dB signal to noise ratio, and separate power supplies for audio and digital. Weighs almost 30 pounds, originally cost $2000.
Unfortunately, I was not as lucky with this second purchase of a 5900 as I was with the first which went into my living room. The second was not as clean looking, and has a noticeable whine on high speed discs like DVD-Audio and SACD. It may need internal cleaning and/or a new bearing or drive mechanism. Mind you, the unit works OK, it plays every disc I throw at it, it's just more mechanically noisy than a top shelf unit should be.
I figured I might be able to get the noise fixed at the Denon service center about 100 miles away at probable cost $100-$250.
But I had a cheaper solution. I simply avoided retiring my previous bedroom player, a Denon 2900. Very similar to the 5900 except lacking the dual power supplies and also lacking HDCD. It uses the early-production version of the same DAC as the 5900. This 2900 I bought a few years ago is minty clean.
So what I do is this: I play low speed discs, CD's and HDCD's, on the 5900, and I play high speed discs, DVD-Audio and SACD, on the 2900. Only the 2900 needs a video connection, and it connects to video monitor using S-Video through a isolation transformer to prevent contaminating the audio.
This is only for actual discs. Mostly I listed to computer files through my Sonos system. But Sonos doesn't handle high resolution formats, and I don't necessarily have all my discs copied to Sonos (copying discs to an online library is an unbelievable amount of work if you are somewhat perfectionistic about it).
In the long run, I have two nearly identical machines, so one could serve as spare parts for the other. I've even thought about transferring the quiet mechanism from the 2900 to the 5900, but for the time being, why bother?
Both players feed analog into my digitally processed system. For many years, I have used a TacT RCS 2.0 preamp in the bedroom, which feeds Behringer DEQ 2496 and DCX 2496. Unfortuantely, the analog input on the Tact features a somewhat outdated analog-to-digital converter which effectively only has about 16 bit performance.
But late last year I bought a new analog-to-digital converter (ADC). It's actually another Behringer DEQ 2496, which I plan to use for other purposes also, but inside the DEQ is a fairly decent AKM ADC.
I've had this hooked up, but not very well until last night. Previously I ran the Denon players through my Acurus L10 preamp to set the level, then into the DEQ. But last night I decided to bypass the gain stage and volume control of the Acurus, and use the tape output instead. That seemed to really open it up, just like the living room system when I play advanced resolution discs.
And of all the kooky things I was not connecting the Behringer "ADC" to my Tact very well. I was using the optical output of the Behringer. But then it had to share the optical input on the Tact (which has only one optical input) with my carousel DVD player. Last night I got tired of switching the optical cables, and decided to connect the Behringer using AES/EBU instead.
Wednesday, January 26, 2011
Better Highs with Better Bias
One of the ideas in last post was to re-bias Acurus A250 amplifier which currently drives supertweeters. Here's a discussion on exactly this sort of thing with a virtually identical Acurus A200, where it is claimed by several posters that increased bias removes the steely highs of this amplifier:
http://www.diyaudio.com/forums/solid-state/157213-acurus-a200-schematic-here.html
Increasing bias to 6-7mA (measured across 0.5 ohm emitter resistors) is said to improve highs immensely.
It's not entirely clear to me if the R11 potentiometer is the bias adjustment; that's the only potentiometer I see on the schematic. Is there no offset adjustment then?
http://www.diyaudio.com/forums/solid-state/157213-acurus-a200-schematic-here.html
Increasing bias to 6-7mA (measured across 0.5 ohm emitter resistors) is said to improve highs immensely.
It's not entirely clear to me if the R11 potentiometer is the bias adjustment; that's the only potentiometer I see on the schematic. Is there no offset adjustment then?
Supertweeter integration and buyer's remorse
Listening to a very hot (in many ways) brass band recording called "Hot House" by Arturo Sandeval, it seemed like my Elac 4pi supertweeters were sounding metallic. Of course, the recording itself is very brassy, and very bright, so perhaps it's not surprising that the supertweeters are adding the brassy sound that the Acoustats are unable to reproduce. Listening to the supertweeter(s) (or just one) by itself, it was like pure brassiness. Way over the top.
This has begun to bug me a bit. My current supertweeter integration, with much higher supertweeter level and lower crossover point than I used last year, still leaves something to be desired.
One sacrifice I have already decided to make. I might have liked to sit closer to the speakers for a wider stereo image, but can't because the supertweeters begin to become prominently noticeable as a sound source.* So then I found a new compromise position 10 inches back, about 6 feet from the speakers, which is much better on that score (and six feet back from the speakers is probably a good idea anyway). But still not perfect. I still do feel I can hear the supertweeters as a separate source on some music. Sometimes what I seem to hear is that the acoustats have a bump on the side where the supertweeter is, and it is this whole oddly shaped conglomeration that is the speaker. I suppose ideally I wouldn't hear any speaker at all, and some recordings can pull off that trick.
(*The correct solution for wider stereo image would not be to try to sit closer than 6 feet from speakers. That is not good for most speakers anyway. The correct solution would be to move the speakers more than 5.5 feet apart. Unfortunately, room geometry and multi-purposing does not permit that. The speakers are already as wide apart as I can practically get them, except possibly my moving closer to the wall, but then that wouldn't increase the stereo angle either.)
I also heard some brassiness on Computer World by Kraaftwork, though it varied from track to track. I believe this recording is 100% synthesized, shouldn't sound very brassy.
For what it's worth, I've come up with a new compromise cutoff point for the supertweeter, 15.5kHz. The old "magic" 14.2 is just way over the top on a brassy recording. The refined (and accurate by some measurements) 16.5kHz just doesn't have enough supertweeter magic. I previously argued that 14.8kHz was the top of the magic range, but have decided that 15.5kHz has just enough trickery to still be fun.
Interesting that lots of supertweeter users never even bother to set the supertweeter level. For many supertweeters, you simply set the cutoff frequency to make the supertweeter unobtrusive. Supertweeters are generally designed to be more efficient than most loudspeakers, so following the cutoff-only adjustment will lead to excess output, but the heightened response above hearing range is described as a "feature" making up for hearing loss.
In my case, I'm following the apparent guidance of two ancient uncalibrated Class A SPL meters, and setting the supertweeter level so that the range 18-24kHz is, as best I can measure, essentially the same as the level at 1kHz. It's possible my meters are so old they are not showing as much high frequency output as they should, so actually I am adjusting the level far higher than it should be (which is what my iPhone RTA app says).
But, for now I'm sticking with the +15dB level setting on the Behringer crossover, attenuated with a 6dB attenuator at the input of the Acurus amplifier. That seems to be the correct level, and the one that can provide the magic supertweeter effect (including, for example, tighter bass).
I'm thinking perhaps the sound could be improved with a better amplifier on the Elacs, like a Class A transistor or tube amp. Or perhaps just re-biasing the current Acurus A250 amplifier. Old Acurus amplifiers are notorious for developing a low bias condition. The bias was already set pretty low (the amp runs cold at idle) compared with sister upline Aragon amplifiers. The Aragon amps differed partly by having about 2-3x the amount of bias. So if the already low bias of the Acurus gets lowered even more through part ageing, you start to get noticeable crossover notch distortion. Such distortion is especially noticeable around the smallest voltage swings, close to 0 volts output, exactly what is required to drive a supertweeter. So I mean to put this on my to-do list for real-soon-now: rebias the Acurus amp, possibly to slightly higher than factory bias. I could also try one of my other spare amplifiers; the Marantz 15 and Parasound HCA-1000A are good possibilities. Unfortunately, my McIntosh MC225, a 25 watt per channel tube amp with high bias Class AB operation and unity coupled output, needs to be refurbished, but it might be a great amp for this amplification.
One thing I barely think of at all, and that is putting lowpass crossover on the Acoustats above 14kHz or even lower, to help suppress their natural resonance around 13.8kHz. I just don't want to do it, but it might actually help. Despite protestation to the contrary, I am still relying on TacT RCS system to do the final frequency balancing, which can't really be done entirely with crossover hacking anyway.
I believe it was Harry Pearson (famous Founder of The Absolute Sound) who said the Acoustats pull off a trick. They make up for lack of extreme highs (above 14kHz) by having somewhat elevated response below that in the 10-13khz range. I am trying to undo this compromise with a supertweeter (something Harry, IIRC, suggested that the Acoustat designer James Strickland should look into) giving me fully extended response. It follows that, at some point I need to remove the 10-14kHz elevation.
A curious thing about the Elacs is that it does not seem to radiate most sound from the center of any face. It seems to radiate from the edges. This may result from the unique kind of sawtooth aluminum ribbon that Elac uses. It almost seems that it might operate like a Heil tweeter, but all the information I have suggests that it is a pure aluminum ribbon operating in a circular magnetic field. The sawtooth construction allows the ribbon to expand in a linear fashion, which is a great idea. But it may also cause curious radiation patterns, even potential resonances. Put right next to another speaker, it extends the radiating pattern farther to the side than you might predict.
I might get some benefit putting a damping foam wedge between the Elac tweeters and the Acoustat sides which are only a few inches apart. Strangely, I do not seem to hear anything like diffraction around that edge. I can hear the supertweeter perfectly well on the edge across the front of the Acoustat grille...incredible! But when I put my hand, curved, in between the speakers I though it took away some of the metallic sound.
Finally, one thing which has really begun to spark my curiosity comes from measurements I made a few days ago. Achieving what looked like a 14kHz acoustic crossover on the Elac supertweeters required
a 16.5kHz crossover setting on the Behringer 2496 DCX crossover. This was using 48dB Linkwitz-Riley crossover setting (LR48). Given that the response of the Elacs is quite flat from 8kHz to 35kHz (though I'm having trouble finding a graph right now online), setting the crossover that high should not have been necessary.
That makes me wonder a lot if the built-in crossover of the supertweters, which is something like 12dB/octave at 8kHz, is actually detuning the LR48 crossover somehow, and resulting in less than optimal highpass performance at 14kHz.
I keep thinking about this, and on the one hand it seems impossible and other the other hand quite plausible. And if, indeed, the steep electronic crossover is being made less steep because of a passive network attached to the speaker, what should be done about it? (Removing the passive network inside the Elacs would probably not be a good idea for safety reasons...it protects the driver from damage which is quite possible inside my madhouse).
This has begun to bug me a bit. My current supertweeter integration, with much higher supertweeter level and lower crossover point than I used last year, still leaves something to be desired.
One sacrifice I have already decided to make. I might have liked to sit closer to the speakers for a wider stereo image, but can't because the supertweeters begin to become prominently noticeable as a sound source.* So then I found a new compromise position 10 inches back, about 6 feet from the speakers, which is much better on that score (and six feet back from the speakers is probably a good idea anyway). But still not perfect. I still do feel I can hear the supertweeters as a separate source on some music. Sometimes what I seem to hear is that the acoustats have a bump on the side where the supertweeter is, and it is this whole oddly shaped conglomeration that is the speaker. I suppose ideally I wouldn't hear any speaker at all, and some recordings can pull off that trick.
(*The correct solution for wider stereo image would not be to try to sit closer than 6 feet from speakers. That is not good for most speakers anyway. The correct solution would be to move the speakers more than 5.5 feet apart. Unfortunately, room geometry and multi-purposing does not permit that. The speakers are already as wide apart as I can practically get them, except possibly my moving closer to the wall, but then that wouldn't increase the stereo angle either.)
I also heard some brassiness on Computer World by Kraaftwork, though it varied from track to track. I believe this recording is 100% synthesized, shouldn't sound very brassy.
For what it's worth, I've come up with a new compromise cutoff point for the supertweeter, 15.5kHz. The old "magic" 14.2 is just way over the top on a brassy recording. The refined (and accurate by some measurements) 16.5kHz just doesn't have enough supertweeter magic. I previously argued that 14.8kHz was the top of the magic range, but have decided that 15.5kHz has just enough trickery to still be fun.
Interesting that lots of supertweeter users never even bother to set the supertweeter level. For many supertweeters, you simply set the cutoff frequency to make the supertweeter unobtrusive. Supertweeters are generally designed to be more efficient than most loudspeakers, so following the cutoff-only adjustment will lead to excess output, but the heightened response above hearing range is described as a "feature" making up for hearing loss.
In my case, I'm following the apparent guidance of two ancient uncalibrated Class A SPL meters, and setting the supertweeter level so that the range 18-24kHz is, as best I can measure, essentially the same as the level at 1kHz. It's possible my meters are so old they are not showing as much high frequency output as they should, so actually I am adjusting the level far higher than it should be (which is what my iPhone RTA app says).
But, for now I'm sticking with the +15dB level setting on the Behringer crossover, attenuated with a 6dB attenuator at the input of the Acurus amplifier. That seems to be the correct level, and the one that can provide the magic supertweeter effect (including, for example, tighter bass).
I'm thinking perhaps the sound could be improved with a better amplifier on the Elacs, like a Class A transistor or tube amp. Or perhaps just re-biasing the current Acurus A250 amplifier. Old Acurus amplifiers are notorious for developing a low bias condition. The bias was already set pretty low (the amp runs cold at idle) compared with sister upline Aragon amplifiers. The Aragon amps differed partly by having about 2-3x the amount of bias. So if the already low bias of the Acurus gets lowered even more through part ageing, you start to get noticeable crossover notch distortion. Such distortion is especially noticeable around the smallest voltage swings, close to 0 volts output, exactly what is required to drive a supertweeter. So I mean to put this on my to-do list for real-soon-now: rebias the Acurus amp, possibly to slightly higher than factory bias. I could also try one of my other spare amplifiers; the Marantz 15 and Parasound HCA-1000A are good possibilities. Unfortunately, my McIntosh MC225, a 25 watt per channel tube amp with high bias Class AB operation and unity coupled output, needs to be refurbished, but it might be a great amp for this amplification.
One thing I barely think of at all, and that is putting lowpass crossover on the Acoustats above 14kHz or even lower, to help suppress their natural resonance around 13.8kHz. I just don't want to do it, but it might actually help. Despite protestation to the contrary, I am still relying on TacT RCS system to do the final frequency balancing, which can't really be done entirely with crossover hacking anyway.
I believe it was Harry Pearson (famous Founder of The Absolute Sound) who said the Acoustats pull off a trick. They make up for lack of extreme highs (above 14kHz) by having somewhat elevated response below that in the 10-13khz range. I am trying to undo this compromise with a supertweeter (something Harry, IIRC, suggested that the Acoustat designer James Strickland should look into) giving me fully extended response. It follows that, at some point I need to remove the 10-14kHz elevation.
A curious thing about the Elacs is that it does not seem to radiate most sound from the center of any face. It seems to radiate from the edges. This may result from the unique kind of sawtooth aluminum ribbon that Elac uses. It almost seems that it might operate like a Heil tweeter, but all the information I have suggests that it is a pure aluminum ribbon operating in a circular magnetic field. The sawtooth construction allows the ribbon to expand in a linear fashion, which is a great idea. But it may also cause curious radiation patterns, even potential resonances. Put right next to another speaker, it extends the radiating pattern farther to the side than you might predict.
I might get some benefit putting a damping foam wedge between the Elac tweeters and the Acoustat sides which are only a few inches apart. Strangely, I do not seem to hear anything like diffraction around that edge. I can hear the supertweeter perfectly well on the edge across the front of the Acoustat grille...incredible! But when I put my hand, curved, in between the speakers I though it took away some of the metallic sound.
Finally, one thing which has really begun to spark my curiosity comes from measurements I made a few days ago. Achieving what looked like a 14kHz acoustic crossover on the Elac supertweeters required
a 16.5kHz crossover setting on the Behringer 2496 DCX crossover. This was using 48dB Linkwitz-Riley crossover setting (LR48). Given that the response of the Elacs is quite flat from 8kHz to 35kHz (though I'm having trouble finding a graph right now online), setting the crossover that high should not have been necessary.
That makes me wonder a lot if the built-in crossover of the supertweters, which is something like 12dB/octave at 8kHz, is actually detuning the LR48 crossover somehow, and resulting in less than optimal highpass performance at 14kHz.
I keep thinking about this, and on the one hand it seems impossible and other the other hand quite plausible. And if, indeed, the steep electronic crossover is being made less steep because of a passive network attached to the speaker, what should be done about it? (Removing the passive network inside the Elacs would probably not be a good idea for safety reasons...it protects the driver from damage which is quite possible inside my madhouse).
Monday, January 24, 2011
Chair position adjustment and The Near Field
While it was a big improvement in audio realism to move my new listening chair from the far wall to the center of the room last year, I had some difficulty determining how far back into the room (from the speakers) to place the chair. I ended up with a kind of compromise position. Compared to other systems I have heard, it does not give me the widest possible separation. I think it's just shy of 60 degrees angle between the speakers, perhaps 55 degrees. But it is way more separation than I had been accustomed to, having previously listened only from the back of the room.
But it seemed to me that when I placed the speakers even closer to the speakers, the center image fell apart, and that was the main reason for not positioning the listening chair even closer to the speakers.
I've also had a longstanding problem with the center image (usually the lead vocal) not being exactly in the center. When I notice the lead being off to the side in some way not intended, it bothers me a lot. During the past year, I've found that speaker positioning to the nearest 5mm, or delay adjustments as small as 0.03 milliseconds can make a difference in center image stability. But the best thing to insure center image stability is (1) exact chair and speaker positioning, and (2) subwoofer levels. For some reason, the subwoofer levels have an uncanny effect on the center image.
So now that's I've got those problems "solved", and since I am still procrastinating on doing a new Room Correction with Tact, I thought I'd look at the chair position issue again.
I tried a position about 10 inches closer to the front of the room (I define "front" as where the speakers are) than my current position. It did have a noticeably (though not compellingly) wider stereo image. And center image position and stability (assessed by playing Supertramp Crime of the Century) seemed OK. But now what seems irksome is that it appears possible to "hear" the supertweeters as a distinct source. My new much higher supertweeter levels contribute to that.
I tried fiddling with the supertweeter crossover adjustments to make it work better. Of course, I had to increase time delay because of the angles involved. But that wasn't enough. I raised the crossover point. It helped if I raised crossover up to 14.8kKz, but that still didn't make the supertweeter disappear. 16.5kHz did make the supertweeter invisible, but then I loose all the new WOW factor of my fully extended highs which I am really beginning to appreciate.
What is happening is that when I get too close to the speaker (this position was 5 feet from the panels) I start listening from something like "the near field" which tends to exaggerate highs and bass. The room acoustics interact little and are relatively unimportant. In some ways this is good, but most recordings and speakers aren't designed to sound good this way.
So I went back to my "old" position, or really a new "old" position that is still about 5 inches closer to the speakers than last year. It is indicated by my ears lining up with the back edge of the largest side bookcase. This is about 6 feet back from the speakers. Last year I had my ears line up with the forward edge of the next bookcase.
Now I have completed the preliminaries to Room Correction. I have set up low and high frequency crossovers and chosen a listening position. System is sounding very good even without correction, though slightly weak in the 100-500Hz region, lumpy in the deep bass around 45Hz, and just a little too prominent in the upper highs starting at 10K. Those are exactly the kinds of issues that can be solved with room correction. What my RCS 2.0 system cannot do is set up the crossover optimally, but that work is now done, at least good enough for this time around.
But it seemed to me that when I placed the speakers even closer to the speakers, the center image fell apart, and that was the main reason for not positioning the listening chair even closer to the speakers.
I've also had a longstanding problem with the center image (usually the lead vocal) not being exactly in the center. When I notice the lead being off to the side in some way not intended, it bothers me a lot. During the past year, I've found that speaker positioning to the nearest 5mm, or delay adjustments as small as 0.03 milliseconds can make a difference in center image stability. But the best thing to insure center image stability is (1) exact chair and speaker positioning, and (2) subwoofer levels. For some reason, the subwoofer levels have an uncanny effect on the center image.
So now that's I've got those problems "solved", and since I am still procrastinating on doing a new Room Correction with Tact, I thought I'd look at the chair position issue again.
I tried a position about 10 inches closer to the front of the room (I define "front" as where the speakers are) than my current position. It did have a noticeably (though not compellingly) wider stereo image. And center image position and stability (assessed by playing Supertramp Crime of the Century) seemed OK. But now what seems irksome is that it appears possible to "hear" the supertweeters as a distinct source. My new much higher supertweeter levels contribute to that.
I tried fiddling with the supertweeter crossover adjustments to make it work better. Of course, I had to increase time delay because of the angles involved. But that wasn't enough. I raised the crossover point. It helped if I raised crossover up to 14.8kKz, but that still didn't make the supertweeter disappear. 16.5kHz did make the supertweeter invisible, but then I loose all the new WOW factor of my fully extended highs which I am really beginning to appreciate.
What is happening is that when I get too close to the speaker (this position was 5 feet from the panels) I start listening from something like "the near field" which tends to exaggerate highs and bass. The room acoustics interact little and are relatively unimportant. In some ways this is good, but most recordings and speakers aren't designed to sound good this way.
So I went back to my "old" position, or really a new "old" position that is still about 5 inches closer to the speakers than last year. It is indicated by my ears lining up with the back edge of the largest side bookcase. This is about 6 feet back from the speakers. Last year I had my ears line up with the forward edge of the next bookcase.
Now I have completed the preliminaries to Room Correction. I have set up low and high frequency crossovers and chosen a listening position. System is sounding very good even without correction, though slightly weak in the 100-500Hz region, lumpy in the deep bass around 45Hz, and just a little too prominent in the upper highs starting at 10K. Those are exactly the kinds of issues that can be solved with room correction. What my RCS 2.0 system cannot do is set up the crossover optimally, but that work is now done, at least good enough for this time around.
Saturday, January 22, 2011
14.2kHz highpass brings back the magic ultraviolet
Well, all the measurements that led to setting the crossover up to 16.5kHz were based on this assumption: we want to cross the drivers over, not blend them. One driver should be the main contributor, or the other one, thus maximum frequency specificity.
It's a good idea, anyway. But in listening to Bass Erotica and We Want To Be Loved, it seemed to me the magic I heard the other night, the magic was gone, even replaced by a barely noticable harshness (making you think the answer would be to turn the high pass up even more, get them tweeters outa there).
So I went back to 14.2, and the magic was back. It was pretty good at 14.8, quite good at 14.5, and virtually unlistenable at 13.9. At 12kHz, it was intolerable. So 14.8-14.2 seems to be the sweet spot, with 14.2 having the most magic.
Doing a sweep, it does not appear that the Elac supertweeters are contributing much to the peaks of the Acoustat below 14kHz. At 13.9 kHz specifically, turning off the Elacs only reduced the level of a peak by 1dB.
But played by themselves, the Elacs would now be acceptable with another driver crossed over as low as 12khz, according to previous tests. So the current crossover seems to be more of a "blending", with the Elac and Acoustats complementing each other in the range 12-14kHz.
Strangely, however, if I could make the Elacs acoustically cross exactly where I want, I would choose 14.2, and that is also what I have dialed in, but based on listening tests rather than measurements.
An overall sweep shows barely any difference, perhaps the 14.2 provides somewhat greater stability and non-directionality thanks to greater coverage by the omndirectional tweeter rather than the narrowing beam of an electrostatic driver at its upper limits.
It's a good idea, anyway. But in listening to Bass Erotica and We Want To Be Loved, it seemed to me the magic I heard the other night, the magic was gone, even replaced by a barely noticable harshness (making you think the answer would be to turn the high pass up even more, get them tweeters outa there).
So I went back to 14.2, and the magic was back. It was pretty good at 14.8, quite good at 14.5, and virtually unlistenable at 13.9. At 12kHz, it was intolerable. So 14.8-14.2 seems to be the sweet spot, with 14.2 having the most magic.
Doing a sweep, it does not appear that the Elac supertweeters are contributing much to the peaks of the Acoustat below 14kHz. At 13.9 kHz specifically, turning off the Elacs only reduced the level of a peak by 1dB.
But played by themselves, the Elacs would now be acceptable with another driver crossed over as low as 12khz, according to previous tests. So the current crossover seems to be more of a "blending", with the Elac and Acoustats complementing each other in the range 12-14kHz.
Strangely, however, if I could make the Elacs acoustically cross exactly where I want, I would choose 14.2, and that is also what I have dialed in, but based on listening tests rather than measurements.
An overall sweep shows barely any difference, perhaps the 14.2 provides somewhat greater stability and non-directionality thanks to greater coverage by the omndirectional tweeter rather than the narrowing beam of an electrostatic driver at its upper limits.
14-35kHz
And perhaps I shouldn't brag about this, but the extremes in this 14Hz-35kHz are not rolled off, or at least the 14Hz isn't. Output is just as strong at 14Hz as it is at 1Khz, and the subwoofer doesn't even seem to be moving at all. Why I should be bragging is that if I have problems, it is not so much at the extremes, it is in the middle. And all audiophiles know you have to get the middle right first. Well my middle is pretty much all the Acoustat 1+1 speaker, one of the most well regarded speakers in audio history. But by the standards of frequency flatness in my room, it does not at all have flat frequency response. It has too little lower midrange and too much upper midrange. But I am planning, ultimately, to fix that with my Tact RCS 2.0. But even as it is now, with considerable variation in the middle range frequency response, it sounds very very good.
And I am running those supertweeters from 14kHz to 35kHz. (So now you can see the title of this blog entry is also a pun.)
Last night's liveblogging gave you a feel for what would have been 10 minutes of fiddling if I hadn't been trying to blog also, which turned it into an incredibly tedious 45 minutes. It also gave you a feel for how inconsistencies are constantly popping up and you have to decide which ones to investigate. That's just the way it is being an Audio Investigator.
But what I've done since last night is to fully sweep out the left channel (which looks just as good or better than the right...goes down 1Hz lower!). And get around to re-adjusting the supertweeter crossover, which hadn't been done after I raised the supertweeter level from +10 to +15 on the Behringer crossover.
The intent, all along, has been smooth transition between drivers, with no bulge or notch at the crossover. On the supertweeter, I am not lowpassing the Acoustat panels, I am letting them naturally roll off. I thought they naturally rolled off around 18kHz, but my measurements are making it look more like 14kHz. They certainly begin to take a dive at 14kHz, and I found some other measurements on the web to confirm that. In the lower 14kHz range, a serious drop out begins unless you start filling it in with the supertweeter, which is what (at great effort) I have accomplished.
So my intent is to begin the supertweeter at 14kHz. Previously, at the +10dB level, I had decided the correct crossover point was 14.8kHz. Now, in the right (other) channel, and +15dB level, it seems the output doesn't drop -6dB (as would be required by Linkwitz-Riley crossover type) until 12kHz. After many minutes of fiddling and measuring (with the meter needle swinging wildly because of the pure sine waves I am using because I don't have a warble generator) I found that 16.5kHz was required to make the supertweeters have a 14kHz acoustic cutoff (so the nice "filling" described in the above paragraph actually happens, without contributing to peaks caused by the Acoustat below 14kHz...or at least contributing more than a tiny amount).
Now why it is necessary to adjust the crossover to 16.5 kHz (as high or higher than I set it last year...before the supertweeter religion really set in) when what I want is a 14kHz crossover? I often (but incorrectly) assume that the combination of the supertweeter's own built-in crossover (at 8kHz or so) and my electronic line-level crossover (which is in cascade with no interaction) would raise the crossover point, so if I wanted 14kHz I'd need to set 12kHz (?)
Now I think it works the other way, perhaps the supertweeter's built-in crossover, if it is having any significant effect at all, would be to LOWER the high pass cutoff from whatever I set. So it could actually be (though I doubt this is the entire explanation) that I would need to set a 16.5kHz cutoff to actually get 14kHz, just as I have done.
Or maybe it's something else that I haven't thought of yet.
But that's the way these things often are. Anyway, my measurements suggest that at the +15dB level, I need a 16.5kHz cutoff to get nice separation between the Acoustat panels and the Elac supertweeters at 14kHz.
I'm hoping this won't kill the magic time travel I've been experiencing through audio since I boosted the supertweeters to +15dB a few days ago.
(BTW, clarification of earlier post, supertweeter amplifier now has a 6dB attenuator.)
Liveblogging the sweep
FFT spectra are usually crude. Even many apparent continuous displayed spectra are often based on small sample of some kind.
To really get down to the bare metal, you need to turn the oscillator knob by hand and take readings. Unfortunately, this is far easier said than done. At any given precise frequency, the level may fluctuate over a 20dB range just by moving a half meter or so. So which is the "correct" reading? (Here I can see, a warble tone generator would be very very nice...)
Anyway, I got the B&K Oscillator (cheap ugly but working) and Genrad 1933 meter (cost about as much as a car when new around 1973). I've done all my previous sweeping and tuning on the right channel. Now it's time to verify that I haven't messed things up too badly in the left.
Right now I'm playing 14Hz at the minimum output level of the B&K at gain 95 on the Tact.
14 Hz 80dB (effortless. speaker hardly moving)
16 Hz 85dB I can hear it
15 Hz 84dB barely audible
17 Hz 86dB
19-20 Hz 89dB some rattling, chuffing, or something
18 Hz 87dB
21 Hz 94dB vague noise, too loud
22 Hz 93dB too loud
22-23 91
25 89
30 85 sounds plenty loud
45 90 local peak
71 85 local peak
58 80 min, but similar all around here
78 60dB (ok, i tuned this by had to minimize the meter)
89 91
To really get down to the bare metal, you need to turn the oscillator knob by hand and take readings. Unfortunately, this is far easier said than done. At any given precise frequency, the level may fluctuate over a 20dB range just by moving a half meter or so. So which is the "correct" reading? (Here I can see, a warble tone generator would be very very nice...)
Anyway, I got the B&K Oscillator (cheap ugly but working) and Genrad 1933 meter (cost about as much as a car when new around 1973). I've done all my previous sweeping and tuning on the right channel. Now it's time to verify that I haven't messed things up too badly in the left.
Right now I'm playing 14Hz at the minimum output level of the B&K at gain 95 on the Tact.
14 Hz 80dB (effortless. speaker hardly moving)
16 Hz 85dB I can hear it
15 Hz 84dB barely audible
17 Hz 86dB
19-20 Hz 89dB some rattling, chuffing, or something
18 Hz 87dB
21 Hz 94dB vague noise, too loud
22 Hz 93dB too loud
22-23 91
25 89
30 85 sounds plenty loud
45 90 local peak
71 85 local peak
58 80 min, but similar all around here
78 60dB (ok, i tuned this by had to minimize the meter)
89 91
Friday, January 21, 2011
More Crossover Tests and Thoughts
Thursday night I tested my new high supertweeter output level using a General Radio 1933 Sound Analyzer, and it suggests I am doing OK.
The Genrad has built-in switchable octave band filters. I played pink noise and tried the different bands, pointing the random incidence microphone toward the ceiling mostly. The highest band, 16Khz, was pretty much in line with the other bands. If I turn off the supertweeter, the 16Khz band goes way down.
In the first test i did, the 16Khz band was 2dB higher than the 8kHz band. I could easily make it level with the 8kHz band by turning down the supertweeter level a few dB. But then in a later measurement, with the microphone pointed slightly differently, the 16kHz band was 1dB lower than the 8Khz band. The highest bands were the 32Hz and 64Hz bands, they were at least 5dB higher than the others. (Those bands are still in need of room correction with my Tact preamp, but having them boosted slightly makes for better sound anyway.)
More listening to pop recordings such as "Bass Erotica" suggests the new crossover alignment is really on track if not perfect yet. As currently configured, the system is more fun to listen to than it has ever been. It's good all around (if still a bit lumpy in the bass) but the rhythm, pace, and intertransient silence have never been equaled in any of my previous systems.
I particularly think the idea of spacing the low frequency crossover around the 90-100Hz vertical room resonance is fabulous. You are going to have that resonance anyway, so you might as well take advantage of it to take some load off your speakers. Separating the low pass and high pass crossover points reduces the effect of interference that can occur when both speakers are playing simultaneously.
During Crossover Redesign Days 2-3 I spent much time sweeping the low end with my B&K oscillator, and polished the crossover points slightly, moving the subwoofer lowpass from 72 to 84 (to reduce a depression in that range) and moving the panel highpass from 100Hz to 104Hz (because there was still a peak at 100, but when I raised highpass to 107Hz I start getting a depression).
I might try fiddling with this some more...I could possibly eliminate any peaking around 90Hz entirely, but since I am getting to the point where depressions open up if I lower the lowpass or raise the highpass, it would seem I may have already gotten about as much benefit from this trick as I can. And it does nothing for the main room resonance at 45Hz, that will have to be fixed by the Room Correction System.
I've tried different crossover points for the supertweeter also, currently 14.8khz seems to work best (at least when I tested it at the old +10db level).
One concern has also come up in my mind. It seems pretty clear the Acoustats are rolling off pretty steeply above 14kHz. I had thought they were supposed to have high frequency response to 18 or 22kHz. Perhaps mine need some capacitor replacement.
The Genrad has built-in switchable octave band filters. I played pink noise and tried the different bands, pointing the random incidence microphone toward the ceiling mostly. The highest band, 16Khz, was pretty much in line with the other bands. If I turn off the supertweeter, the 16Khz band goes way down.
In the first test i did, the 16Khz band was 2dB higher than the 8kHz band. I could easily make it level with the 8kHz band by turning down the supertweeter level a few dB. But then in a later measurement, with the microphone pointed slightly differently, the 16kHz band was 1dB lower than the 8Khz band. The highest bands were the 32Hz and 64Hz bands, they were at least 5dB higher than the others. (Those bands are still in need of room correction with my Tact preamp, but having them boosted slightly makes for better sound anyway.)
More listening to pop recordings such as "Bass Erotica" suggests the new crossover alignment is really on track if not perfect yet. As currently configured, the system is more fun to listen to than it has ever been. It's good all around (if still a bit lumpy in the bass) but the rhythm, pace, and intertransient silence have never been equaled in any of my previous systems.
I particularly think the idea of spacing the low frequency crossover around the 90-100Hz vertical room resonance is fabulous. You are going to have that resonance anyway, so you might as well take advantage of it to take some load off your speakers. Separating the low pass and high pass crossover points reduces the effect of interference that can occur when both speakers are playing simultaneously.
During Crossover Redesign Days 2-3 I spent much time sweeping the low end with my B&K oscillator, and polished the crossover points slightly, moving the subwoofer lowpass from 72 to 84 (to reduce a depression in that range) and moving the panel highpass from 100Hz to 104Hz (because there was still a peak at 100, but when I raised highpass to 107Hz I start getting a depression).
I might try fiddling with this some more...I could possibly eliminate any peaking around 90Hz entirely, but since I am getting to the point where depressions open up if I lower the lowpass or raise the highpass, it would seem I may have already gotten about as much benefit from this trick as I can. And it does nothing for the main room resonance at 45Hz, that will have to be fixed by the Room Correction System.
I've tried different crossover points for the supertweeter also, currently 14.8khz seems to work best (at least when I tested it at the old +10db level).
One concern has also come up in my mind. It seems pretty clear the Acoustats are rolling off pretty steeply above 14kHz. I had thought they were supposed to have high frequency response to 18 or 22kHz. Perhaps mine need some capacitor replacement.
Thursday, January 20, 2011
Twice as real*...setting the supertweeter level
I believe this is true even if you don't think you can hear that high; I believe that human auditory processing uses higher frequencies even if it seems that we can't hear them. There is published research that shows we are affected by inaudible high frequencies.
After resetting my supertweeter level from +10dB to +15dB as some measurements suggested I should do, I had a glorious night of listening to pop and rock recordings such as Grouse "We Want to be Loved" and Klaatu "Hope" and Genesis "Ripples" (I never really dug that one before but pulled it up by accident) and "Entangled". The clarity, high definition, impactful bass, rhythm, intertransient silence, palpability, liquidity, and continuousness were incredible. Background words were clear as day, bells and other metal objects sounded real and not harsh or overwhelming.
Just as a friend of mine has said, good highs will do more good for the bass than anything. Also, extended extreme highs generally do not cause "harsh" sound, they actually generally facilitate "smoother" less fatiguing sound. It had previously been similarly exciting when I raised the tweeter level from +2dB to +10dB a couple of days ago, I noticed similar effects, but less clearly.
Was it too much, tooooo tooooo much perhaps? Maybe, I'm not sure yet. But I think that if I have not found exactly the correct settings yet, I have found a mark I want to have on the dial, to bring back whenever I want the "twice as real" sensations.
*I once worked for an audio storeowner who often described audio systems...either as criticism or compliment...as "larger than life and twice as real." Whether this was praise of an awesome system or condemnation of an editorializing system depended on the context. In this case, I like my "double reality" so far, but this could change...I've had lots of experiences getting fatigued by high frequency sounds in the past, it's one reason I prefer vintage Acoustat speakers to more modern and obtainable Martin Logan. I have often found Martin Logan speakers to be fatiguing, never so with Acoustats, though the Acoustats conversely seem to lack the nth degree of transparency, which is why I have gotten interested in experimenting with supertweeters in the past two years.
Before getting too concerned about the high supertweeter output level, note that even with this level I was rarely getting one LED on the crossover to light up, and I believe my Elac supertweeters are capable of handling hundreds of watts. Also, to help keep noise down, I have either a 6dB or 10db attenuator on the supertweeter amplifier (an Acurus A250). Can't remember which I'm using right now, but I think it's still the 6dB attenuator. The midrange amplifier level is +3.4, and I think the midrange amplifier has more gain, though the midrange speakers are far less efficient. The Elac tweeters can be crossed over as low as 8Khz, but I'm crossing them at 14.8Khz (plus built-in crossovers). The tweeters are rated to 35Khz and have essentially flat response from 8-35Khz. I think everyone needs good omni ribbon tweeters, and I'm keeping mine, thank you very much, one of my most amazing Audiogon finds. Unfortunately, these days it seems most audiophile are too busy counting their tourmaline cable elevators to bother with supertweeters.
Right now my sound measuring devices are not agreeing on whether this level is correct. My iPhone RTA application is insisting that I've lost my mind, showing 18Khz now about 15dB too high. On the other hand, I have some real but long out-of-calibration Class A sound measuring devices, and they seem to be suggesting if anything not going far enough.
It was using my cracked-glass Bruel & Kjaer 2203 with a 4165 microphone (20Khz free field response) that gave me the inspiration. It seemed to suggest that above their last high frequency resonance around 13.7khz, the Acoustats fade away far too quickly. I tried crossing over the supertweeters at various points, but have now ended up crossing over at 14.8Khz as it seems to prevent the supertweets from contributing significantly to the that last Acoustat resonant peak. With 14.8Khz, there is no serious dip as I roll the oscillator from 13 to 16Khz, yet generally both speakers do not play with the same frequency tone, it's either one or the other, which is very desireable. (However, these tests were done with the +10dB level and now have to be repeated for the +15dB level.)
But whatever level I tried, it didn't seem like I was getting enough output at 16Khz, 20Khz, 22Khz, whatever frequency you care to measure. In that band, where I can still get flat measurements, it seemed like the +10dB setting was still underpowering the supertweets about 6dB below the highs in the 10-13Khz range coming from the Acoustats. Usually when confronted with this sort of data, I seek to split the difference and try something like +12.5dB. But this time, I decided to go all the way that I can go easily, by setting the DCX output level boost to +15 dB, the maximum it allows. That seems to work very well, see subjective comments above, though by measurements with the 2203/4165 combo, I could still add another couple dB to the top end (not so easy to do without changing some equipment) to make it better.
I also have a General Radio meter, one of their last and best ones, and it shows a similar pattern. Surprisingly, the 1" GR electret microphone seems to have even better "high frequency response" than the Bruel & Kjaer, and seems to show I now have flat response to 24Khz. The GR microphone is a random incidence type, not really supposed to be used for measuring speakers (or anything on-axis) but rather backround noise. However, this actually means that, on-axis, it has boosted high frequency response. Used as a random incidence microphone, the GR would only be good to about 16Khz. But it's not clear what the readings actually mean since I am not using this instrument "as directed".
So the instruments don't agree, or I'm not sure exactly what they are saying, but my ears say "don't forget how to do this, it's magic."
It has been incredibly hard to use my Bruel and Kjaer. Possibly my unit itself is flaky (I have two other units that are even worse) and also it turns out to be somewhat difficult to measure super high frequencies in the far field because of beaming, reflections, and interference effects. The meter is constantly moving over a 10dB or greater range as you move the meter slightly or change the frequencies slightly. So it's very hard to pin down my observations to concrete numbers, but as I am scanning the oscillator up and down I can get an idea of where the relative peaks and nulls are.
But that meter does seem to tell me that the acoustats roll off quickly above 14Khz, and I think they shouldn't do that. Possibly they need to have their internal coupling capacitors replaced.
I believe my hearing still goes to 16Khz, but possibly only at quite high levels, I was having trouble hearing above 14.5k from the listening position. But I could clearly tell when the supertweeter was extending the response; with the supertweeter switched on it didn't sound like the response was dropping dead above 14khz, but extends up higher, exactly where is not clear since my hearing is rolled off. Right next to the supertweeter, if i move my head I can hear sonic interference effects up to 16Khz or so.
Wednesday, January 19, 2011
Another Audio Blogger
Somewhere, probably scanning DIYAudio I came across DanTheMan's Blog.
Like me, Dan likes to investigate loudspeakers.
Here's a link to hislist of interesting audio sites
Like me, Dan likes to investigate loudspeakers.
Here's a link to hislist of interesting audio sites
Crossover Redesign, Day One
The 3 day MLK weekend featured cold rainy weather, so I had the perfect alibi to spend 2 days measuring, re-thinking, and re-designing the crossover of my main Living Room System.* This effort is still continuing (now in something like Day Three or more as I am writing this) and I think it has already been extremely beneficial to the sound and will soon make possible a Wonderful New System once I rerun Room Correction.
(*Nominally I tend to use Linkwitz-Riley crossovers whenever possible; they have many advantages starting from the fact that the drivers are always in phase with each other. Currently I use the LR48 slope built into my Behringer DCX 2496 crossover; in addition to being a Linkwitz Riley crossover it is very steep and the drivers are kept in the same polarity. But quite often I adjust crossovers to perform additional frequency countouring by using different crossover frequencies for high and lowpass, etc. This is needed to correct speaker and/or room response, and that is the subject of this post.)
The state of things in late 2010 had clearly gone downhill when I made the measurements (see back a few posts) that showed startling lack of bass response below 100Hz. What happened? I had not only previously measured this as actually boosted below 100Hz but was using a Tact RCS (Room Correction System) 2.0 system to realize my "room curve" which is essentially flat with bass boost below 30Hz. What happened was that over the course of the year, small changes tended to roll back the bass levels. Why were these small changes being made? Because despite reasonable measurements and RCS, the bass response had time smearing properties because of the previous way I had "designed" the crossover between subwoofer(s) and electrostatic panels.
That was waaay back in 2008, I had played pure bass tones from my Kurzweil synthesizer. One of the concerns that I had then was lack of bass in the lowest regions to be reproduced by the panels, in the 100-200Hz range. I used Ivie IE30, Behringer DEQ 2496, and other devices to assess the frequency response and possible ways of setting the crossover. But that was so long ago, I can't really remember much about it except that it proved to me it was OK to use a single subwoofer crossed as high as 86 Hz.
Then later, when I bought the second subwoofer in January 2010, I decided to use higher frequency crossovers to avoid damaging the panels like I had in March 2009 (Acoustat panels cannot arc but you can burn up the transformers with enough sustained high volume levels). To allow me to play as loud as possible with no worries, I pushed the crossover frequency all the way up to 121 Hz. (Previously, I had experimented with crossovers as low as 40Hz.)
By this time, I was already using Tact RCS 2.0 room correction, but I felt I could get the best results out of correction by providing plenty of "material" (sound output) for the RCS algorithms to work with. I would make up for the lack of uncorrected response in the region 100-200Hz by raising the subwoofer lowpass all the way to about 165Hz. So the uncorrected response was now rather "thick" in the bass, but I had no fear about this believing that RCS would fix it up.
Well, RCS did seem to fix it up (though I no longer have any measurements to prove that, wasn't blogging then). But the RCS corrections were apparently very fragile. If you weren't sitting in the correct spot (which moved to mid-room) the bass could get quite boomy. So over time, I turned down the bass, leading to the current situation.
All new work is starting from NO RCS CORRECTION to make it more robust this time. RCS will be added to a good working system to make it even better, rather than being added to a system which merely has enough "material".
Pink noise measurements like the ones in the previous post strongly suggested that the high bass lowpass was not good. When the lowpass was brought down to 121 Hz used by the highpass, response got a lot better.
For several days, I reveled in the idea that maybe I should stick with idealized crossovers, just set the controls so high pass and low pass are the same. But it has passed now, because I found something even better.
Even with low and highpass set to 121 Hz, there was still considerable peaking around 100Hz as shown just below.
So I got to thinking, what if I lowered the lowpass even more. Now this is really going in the opposite direction from what I did last time, but I thought now it might be worth testing.
Sure enough, I can lower the subwoofer lowpass as low as 72Hz and still get seemingly nice response up to 121 Hz where the panels start crossing out. In fact, it is still slightly peaky if I do that, but the peak is lowered to about 80Hz and much smaller.
Uncorrected, this is far better sounding than the straight 121Hz crossover, and has much less peaking in the 100Hz region.
Now some people might complain that I am not crossing over "correctly" because high and low pass frequencies are not set the same. But the proof is in the pudding, I get far flatter response separating the low and high pass. Can I get away with this, and should I?
Yes and yes. It turns out that, very conveniently, speakers and rooms have errors mostly of the type called "minimum phase". If you add a minimum phase error to a minimum phase correction that perfectly cancels it for flat frequency response, it turns out you have also perfectly corrected the phase response as well. (Of course, nothing is ever "perfect" but it can be close.)
Now what is happening in this case? I believe the room has a strong resonant node in the 80-110Hz region caused by ceiling bounce (the ceiling ranges from 8'2" to 9'8"). This has half wave nodes in the 45Hz region, and full wave nodes in the 90Hz region.
A much smaller factor may be that the subwoofers have resonance in the 80-110Hz region. I have not tried to measure such a resonance, but I do sense that the SVS PB-13 is not as perfectly flat to 200Hz as the website suggested.
I currently think the room resonance is the larger factor, and it is very convenient that I can cancel it out by pushing down the lowpass cutoff of the subwoofer even below the highpass cutoff of my panels. I have used this trick previously in other systems, but never documented it as well as I am doing here.
Later in the day, the strange notch frequency around 4-6K began to bother me. But first I asked myself "is this real, or an artifact of the iPhone RTA measuring device." I tried running a spectrum of my bedroom system, based on Revel M20 speakers, which I had previously measured with an Ivie IE30 as being nearly perfectly flat above 1Khz.
Well, that spectrum showed exactly the same notchout at 4-6K, followed by a boost from 8K-13K, followed by a notch, followed by a boost up to 20K. In short, it looked almost exactly the same.
I then checked a spectrum I had run of a friends system (a very committed audiophile) in December. Same features in the response above 4K.
Three systems with the same response above 4K? Not likely, especially as I had previously measured one of those systems as being flat, not what was being shown. I conclude that the iPhone RTA measurement above 4K can't be trusted, or it means more-or-less flat if it shows the pattern shown above.
I think the bass response of RTA is fairly trustworthy, however, and it looked different on each of the three systems tested.
But it is good to beware, even if you like to do objective audio measurements as I do, that any particularly measurement can be, and usually is, flawed in more than one way. That's also why I have many different measurement instruments that do roughly the same thing, because they each reveal different things, and not one of them is perfect. If I had lots more money, I could afford more perfect measuring instruments, they are extremely expensive.
A real time analyzer with lots of bands is one of the most handy tools for setting up an audio system. However, it can also be misleading. It could be misleading even if it had flat response. Frequency coutours of a speaker or room are obscured by the fixed frequency bandwidths. One can get much closer to the actual sound by using a sine wave generator. Plus, with a sine wave generator, I can make spectrum level and continuity judgements by ear, and also use one of many Class A microphones I have to measure sound pressure level.
On day two, I brought out the B&K Function Generator to see what I had done and to make it even better...
(*Nominally I tend to use Linkwitz-Riley crossovers whenever possible; they have many advantages starting from the fact that the drivers are always in phase with each other. Currently I use the LR48 slope built into my Behringer DCX 2496 crossover; in addition to being a Linkwitz Riley crossover it is very steep and the drivers are kept in the same polarity. But quite often I adjust crossovers to perform additional frequency countouring by using different crossover frequencies for high and lowpass, etc. This is needed to correct speaker and/or room response, and that is the subject of this post.)
The state of things in late 2010 had clearly gone downhill when I made the measurements (see back a few posts) that showed startling lack of bass response below 100Hz. What happened? I had not only previously measured this as actually boosted below 100Hz but was using a Tact RCS (Room Correction System) 2.0 system to realize my "room curve" which is essentially flat with bass boost below 30Hz. What happened was that over the course of the year, small changes tended to roll back the bass levels. Why were these small changes being made? Because despite reasonable measurements and RCS, the bass response had time smearing properties because of the previous way I had "designed" the crossover between subwoofer(s) and electrostatic panels.
That was waaay back in 2008, I had played pure bass tones from my Kurzweil synthesizer. One of the concerns that I had then was lack of bass in the lowest regions to be reproduced by the panels, in the 100-200Hz range. I used Ivie IE30, Behringer DEQ 2496, and other devices to assess the frequency response and possible ways of setting the crossover. But that was so long ago, I can't really remember much about it except that it proved to me it was OK to use a single subwoofer crossed as high as 86 Hz.
Then later, when I bought the second subwoofer in January 2010, I decided to use higher frequency crossovers to avoid damaging the panels like I had in March 2009 (Acoustat panels cannot arc but you can burn up the transformers with enough sustained high volume levels). To allow me to play as loud as possible with no worries, I pushed the crossover frequency all the way up to 121 Hz. (Previously, I had experimented with crossovers as low as 40Hz.)
By this time, I was already using Tact RCS 2.0 room correction, but I felt I could get the best results out of correction by providing plenty of "material" (sound output) for the RCS algorithms to work with. I would make up for the lack of uncorrected response in the region 100-200Hz by raising the subwoofer lowpass all the way to about 165Hz. So the uncorrected response was now rather "thick" in the bass, but I had no fear about this believing that RCS would fix it up.
Well, RCS did seem to fix it up (though I no longer have any measurements to prove that, wasn't blogging then). But the RCS corrections were apparently very fragile. If you weren't sitting in the correct spot (which moved to mid-room) the bass could get quite boomy. So over time, I turned down the bass, leading to the current situation.
All new work is starting from NO RCS CORRECTION to make it more robust this time. RCS will be added to a good working system to make it even better, rather than being added to a system which merely has enough "material".
Pink noise measurements like the ones in the previous post strongly suggested that the high bass lowpass was not good. When the lowpass was brought down to 121 Hz used by the highpass, response got a lot better.
For several days, I reveled in the idea that maybe I should stick with idealized crossovers, just set the controls so high pass and low pass are the same. But it has passed now, because I found something even better.
Even with low and highpass set to 121 Hz, there was still considerable peaking around 100Hz as shown just below.
So I got to thinking, what if I lowered the lowpass even more. Now this is really going in the opposite direction from what I did last time, but I thought now it might be worth testing.
Sure enough, I can lower the subwoofer lowpass as low as 72Hz and still get seemingly nice response up to 121 Hz where the panels start crossing out. In fact, it is still slightly peaky if I do that, but the peak is lowered to about 80Hz and much smaller.
Uncorrected, this is far better sounding than the straight 121Hz crossover, and has much less peaking in the 100Hz region.
Now some people might complain that I am not crossing over "correctly" because high and low pass frequencies are not set the same. But the proof is in the pudding, I get far flatter response separating the low and high pass. Can I get away with this, and should I?
Yes and yes. It turns out that, very conveniently, speakers and rooms have errors mostly of the type called "minimum phase". If you add a minimum phase error to a minimum phase correction that perfectly cancels it for flat frequency response, it turns out you have also perfectly corrected the phase response as well. (Of course, nothing is ever "perfect" but it can be close.)
Now what is happening in this case? I believe the room has a strong resonant node in the 80-110Hz region caused by ceiling bounce (the ceiling ranges from 8'2" to 9'8"). This has half wave nodes in the 45Hz region, and full wave nodes in the 90Hz region.
A much smaller factor may be that the subwoofers have resonance in the 80-110Hz region. I have not tried to measure such a resonance, but I do sense that the SVS PB-13 is not as perfectly flat to 200Hz as the website suggested.
I currently think the room resonance is the larger factor, and it is very convenient that I can cancel it out by pushing down the lowpass cutoff of the subwoofer even below the highpass cutoff of my panels. I have used this trick previously in other systems, but never documented it as well as I am doing here.
Later in the day, the strange notch frequency around 4-6K began to bother me. But first I asked myself "is this real, or an artifact of the iPhone RTA measuring device." I tried running a spectrum of my bedroom system, based on Revel M20 speakers, which I had previously measured with an Ivie IE30 as being nearly perfectly flat above 1Khz.
Well, that spectrum showed exactly the same notchout at 4-6K, followed by a boost from 8K-13K, followed by a notch, followed by a boost up to 20K. In short, it looked almost exactly the same.
I then checked a spectrum I had run of a friends system (a very committed audiophile) in December. Same features in the response above 4K.
Three systems with the same response above 4K? Not likely, especially as I had previously measured one of those systems as being flat, not what was being shown. I conclude that the iPhone RTA measurement above 4K can't be trusted, or it means more-or-less flat if it shows the pattern shown above.
I think the bass response of RTA is fairly trustworthy, however, and it looked different on each of the three systems tested.
But it is good to beware, even if you like to do objective audio measurements as I do, that any particularly measurement can be, and usually is, flawed in more than one way. That's also why I have many different measurement instruments that do roughly the same thing, because they each reveal different things, and not one of them is perfect. If I had lots more money, I could afford more perfect measuring instruments, they are extremely expensive.
A real time analyzer with lots of bands is one of the most handy tools for setting up an audio system. However, it can also be misleading. It could be misleading even if it had flat response. Frequency coutours of a speaker or room are obscured by the fixed frequency bandwidths. One can get much closer to the actual sound by using a sine wave generator. Plus, with a sine wave generator, I can make spectrum level and continuity judgements by ear, and also use one of many Class A microphones I have to measure sound pressure level.
On day two, I brought out the B&K Function Generator to see what I had done and to make it even better...
Monday, January 10, 2011
Setting the bass crossover
I improved the response of the right channel from the first spectrum to the second one shown by moving the low pass crossover frequency for the subwoofer from around 160Hz down to 121Hz. The high pass crossover on the electrostatic panels remained the same 121Hz in both cases. Previously I crossed over the subwoofers higher than the panels deliberately in order to fill in what I perceived as a low point in response between 120Hz and 200Hz. In the overall scheme of things, that depression (still visible in second spectrum) doesn't seem as important as the 80-150hz peak that arises from this crude attempt to reduce it, and the bass sounds cleaner (less resonant) with the 121 Hz low pass. Been thinking about pushing crossover even lower, but I don't like to put too much of the low frequency spectrum into the panels as their low frequency transformers can melt, so 121 is about as low as I want to high pass the panels for reliability reasons.
The performance was also improved from previous spectra by increasing bass generally in both channels (through the crossover) and individually in the right channel by setting the volume control on the sub. (I can't remember if these changes got in the first spectrum above, possibly not, though it actually looks like there is some improvement there.)
Also, the supertweeter level was reduced about 5dB, that affects pretty much only the two highest frequency bars, where it is sometimes hard to see any change. The supertweeter is not intended to be very audible, instead just at the margins of audibility, and filling in the upper frequencies which we don't think we hear but might have perceptible effects anyway. Normally, the supertweeters help remove the bandwidth limited "plasticy" sound from the Acoustats; recently they may have been adding some undesireable aggressive steeliness. On other analyzers, it often appears like the supertweeter is improving the response of the highest frequency bar by making it flush rather than lower than the preceding bar, thus extending the high frequency response and making it less peaky.
As the last paragraph makes clear, though spectra look impressive and "scientific" when presented, you have to remember that each measurement is but an imperfect sample of the processes involved, and that better measurements can often put things into a rather different light. Visible and rationally analyzable spectra greatly facilitate progress in audio design, but should never be considered the end goal. The end goal is better sound, and the spectrum is only a tool that helps modify a system towards that goal.
I would classify the frequency response in the second graph as "pretty good", and much better than the average system, but still in need of work for the ultimate level of performance (what I was claiming to provide previously...now again a work in progress since my discovery of low bass in a previous post).
The suckout in the 4K-6K region is troubling, probably deserves more investigation, though most likely is a speaker limitation that can't be cured but can be corrected by the Room Correction System.
Thursday, January 6, 2011
Frequency Response Errors have accumulated
Frequency response is something I consider very important. This system is corrected to be flat with a slight bass boost below 30Hz, through the use of a Tact 2.0 RCS preamp with user room curve settings. I remember pink noise measurements in the past (taken in the same way) which were virtually flat (unfortunately, I have no permanent record of them anymore). The above curve was taken with pink noise from Stereophile Test Disc Two and the iPhone application called RTA by Studio Six Digital.
But not this time. Now I am showing 20dB of response variation from the highest highs to the lowest lows, with mostly a gradually rising response from lows to highs, but very low response in the bass below 100Hz where the subwoofers are playing.
So what happened?
So what happened?
Possibly somehow my Tact has gotten messed up. With the room correction set to bypass, I get more sub 30Hz bass than with correction turned on. Possibly the Tact lost its memory in a power surge, or possibly the changed response was a result of the accumulation of small changes.
Funny how I have made small changes over the year, now ending up with this, clearly not what was intended. I would have not expected much different from flat response. Seems like a quarterly frequency response test is in order to avoid drifting away from desired frequency response through seemingly limited corrections.
Back in mid 2010 when I was playing the Santana "Supernatural" album and concerned about way-too-much-bass, I discovered that right subwoofer level had possibly been tampered with, turned straight up, which was not what I remembered it should be. So I found a reduced level I thought was better, combined with less bass boost used in my room currection curve. I figured 3dB reduction in bass, tops, and I was simply restoring the original corrected response. It sounded much better on Supernatural without any bass cut, and I didn't notice any problem with other recordings later.
To restore frequency response, I will begin with no correction, and set driver levels and crossover points for a fairly decent overall response. Then room correction measurement and room curve adjustment will be performed.
Room correction measurement is a complicated process, and I haven't bothered since early last year.
Subscribe to:
Posts (Atom)