Was that Entec? Amazed that I dredged that name up!
I began thinking about a particular approach when Harman was trying to make the little spiderless inverted dome transducers for laptops etc., which have become widely imitated (a huge mistake not to get a simple design patent on them, but so it goes). They have a rather large ratio of excursion to diameter, and their principal failing, not surprisingly, is IM distortion. The mass is low and the motor huge for the overall size. There is also a nasty resonant peak.
I agree one wouldn't try to do absolute position encoding above some reasonable frequency. Rather, the transition would be to a variable gain that was itself determined in advance, and this applied appropriately to the higher frequency signals.
In principle a "plant model" of the device could be used as well, particularly if the temperature of the voice coil is kept track of, a la Klippel and his mirror filter --- or current drive used, with a safety feature to prevent thermal runaway.
Sean Olive told me he's about to embark on a study of the audibility of speaker nonlinear distortions. Should be interesting. Anecdotally, when Wolfgang K. was shopping his mirror filter around, he presented it to Buck and others at Harman, and because a certain amount of nonlinear distortion tends to make a woofer seem louder, it was not deemed to be something worth pursuing for low-budg speakers!
Of course triangular waves (and not square waves) are particularly good for noticing crossover distortion of a gross sort, i.e., a discontinuity/deadband at the origin, or even an abrupt change of slope for displaced gain changes, on an oscilloscope. They are also good for seeing clipping.
It is true that the easiest-synthesized waveforms from old-fashioned synthesizers will sometimes occasion strange behavior in systems, especially compressor/limiters and companding systems like old Dolby NR. I may have related the story of the UCLA Electronic Music Studio acquiring budget Dolby noise reduction boxes, which were being marketed to amateur recordists. They had cut a few corners and decided that by and large the musical signals being processed were mostly symmetrical. Thus someone decided that a half-wave rectifier was adequate as part of level determination.
The behavior with a sawtooth waveform depended very much on the polarity! In the one case the system worked fine (or as well as Dolby ever does). In the other, it flat-out malfunctioned until the level got quite high. A guest physics grad student and I figured out how to add a circuit in parallel (no cutting of traces on the existing boards, so that in principle the additional parts could be removed) to manage symmetrical response. The downfall was that by removing the boards for the addition, and reinserting, the contacts on the budget edge connectors were fatigued and became intermittent, and failures would occur at the worst-possible times.
I have a little background in high power servos from the old 9-track tape days. This led me to consider servo drivers many times. There are quite a few problems. I have a solution for many, but not the wherewithal or expertise to build it.
First is hysteresis. By my rough calculations, you need to remain within about 10 degrees of phase to where you want to be to a constantly changing target over a pretty wide range. The higher the mass, the higher the force needed. It ain't linear baby! This is the killer. You could do this, but the power required you are not prepared for. It was many years ago when I had the help of a real EE who knew servo, and we calculated to move the Celestion woofer I had on hand at 200 Hz over the same range we expected with my 100W amp, would need in the range of 10,000 W. With class D, this may now be feasible.
Second is measurement that is fast, accurate and in phase. I solved this with my idea of an optical tapered slot and fast digital reading. I prototyped this many years ago for 4 bits using comprators. With delta sigma I believe one could get 12 or so bits within the available timing window.
Third was the look-ahead predictive algorithms. The higher the mass, the more you modify the correction drive to predict it's result, as opposed to trying to play catch up. We pretty much figured out for the wide range of frequency, we could not use analog. At that time, our fast processors were Z80's, so digital was out. I suspect that we could do the calculation with modern chips. It would take a pretty good DSP programmer and it would need to be microcode. Java hacks need not apply.
I came up with a no-mass sensor for a mid or tweet, but the problems of speed and power seem insurmountable even now.
So where are we? Well, it is pretty easy to make a woofer produce a steady tone with very low distortion using simple analog servo. Trouble is, it will be a day late and a dollar short. I have heard some high end servo's and they leave me with a "flat" sense compared to non-servo subs. Building a servo sub may now be possible, but it would sure not be budget. For now, for my cheep budget, I will stick with my critical Q multi subs.
First is hysteresis. By my rough calculations, you need to remain within about 10 degrees of phase to where you want to be to a constantly changing target over a pretty wide range. The higher the mass, the higher the force needed. It ain't linear baby! This is the killer. You could do this, but the power required you are not prepared for. It was many years ago when I had the help of a real EE who knew servo, and we calculated to move the Celestion woofer I had on hand at 200 Hz over the same range we expected with my 100W amp, would need in the range of 10,000 W. With class D, this may now be feasible.
Second is measurement that is fast, accurate and in phase. I solved this with my idea of an optical tapered slot and fast digital reading. I prototyped this many years ago for 4 bits using comprators. With delta sigma I believe one could get 12 or so bits within the available timing window.
Third was the look-ahead predictive algorithms. The higher the mass, the more you modify the correction drive to predict it's result, as opposed to trying to play catch up. We pretty much figured out for the wide range of frequency, we could not use analog. At that time, our fast processors were Z80's, so digital was out. I suspect that we could do the calculation with modern chips. It would take a pretty good DSP programmer and it would need to be microcode. Java hacks need not apply.
I came up with a no-mass sensor for a mid or tweet, but the problems of speed and power seem insurmountable even now.
So where are we? Well, it is pretty easy to make a woofer produce a steady tone with very low distortion using simple analog servo. Trouble is, it will be a day late and a dollar short. I have heard some high end servo's and they leave me with a "flat" sense compared to non-servo subs. Building a servo sub may now be possible, but it would sure not be budget. For now, for my cheep budget, I will stick with my critical Q multi subs.
Here's an idea: Drive the whole speaker cone at once.
Use a very stiff cone, with a surround that is flexible, with a sealed chamber behind it, filled with a low-mass incompressible fluid. Also connected to the fluid chamber would be a piston, the position of which would be controlled by the amplifier output signal.
Use a very stiff cone, with a surround that is flexible, with a sealed chamber behind it, filled with a low-mass incompressible fluid. Also connected to the fluid chamber would be a piston, the position of which would be controlled by the amplifier output signal.
For a bonus, put the (sealed) motor inside the fluid chamber, for better cooling.
The outside face of the piston would probably have to be in room air. So it might try to make some sound. I guess that could be dealt with. Or it could push and pull against the air in an enclosure...
If using triangle-waves to measure the effects of crossover distortion would we expect to see greatly increased 7 th and 9 th harmonics ? Do we need the distortion of the test wave to be -20dB lower ? Maybe even better than that ? That suggest at least - 100 dB to make proper measurements . That being accepted it might be useful .I doubt the triangle wave will show much the sine wave didn't . It would show over or under biasing earlier is my guess ( less than optimum bias with either test method if so ) ?
If you believe a woofer should have a sound most servo woofers will not sound right. I had many customers complain at first about no bass, until they recognized they were expecting a lot of doubling (2nd harmonic) from the woofers. Very few woofers can actually reproduce bass below 60 Hz. Most do a very poor job and actually have lot less output in the low bass than the size would imply.
This is a neat new part that could enable a servo woofer: http://www.ti.com/lit/ds/symlink/ldc1000.pdf similar to the first generation Entec. DSP systems today could manage the process in digital pretty well.
I was never a fan of accelerometer servos. They do the wrong thing when you hit the stops. . .
This is a neat new part that could enable a servo woofer: http://www.ti.com/lit/ds/symlink/ldc1000.pdf similar to the first generation Entec. DSP systems today could manage the process in digital pretty well.
I was never a fan of accelerometer servos. They do the wrong thing when you hit the stops. . .
Triangle waves are fine for large crossover distortion and a simple oscilloscope. Today a spectrum analyzer and a pure source is far easier to use and shows lots more information. The high order harmonics are your enemy.
The mass of the fluid and the speed of sound are hard obstacles to overcome. Incompressible and low mass do not really happen together.
Is it possible to work the drive unit has to be able to create an error signal larger enough to correct ? As far as I can see high damping factor does enough . Acoustic Research were doing all the right things way back in time . Perhaps like aircraft an inherently inferior design lends itself to correction better as in fly by wire stability discussions .
The so called Isobarik loading is worth a thought .
Surely time delay is the big question ? Hi fi adds time delay which is bad . If we need to use digital why not correct this as part of the package ?
The so called Isobarik loading is worth a thought .
Surely time delay is the big question ? Hi fi adds time delay which is bad . If we need to use digital why not correct this as part of the package ?
I went to a demonstration where the presenter re-digitized the output of an amplifier . The output was compared with the original CD as a graph of deviation from the original . Time domain distortion it was said . Great advantage being to use real music . I have heard all the arguments against it . It still seems the right way if reliable and truthful . The demonstration claimed to answer many questions of perception . E.g. disliking something which measurements found to be OK . I have mention this before . This came from the RAF originally . The other RAF thing was NXT . The RAF thought sticking damping panels to the cockpit sides would make it quieter . The opposite happened . One of their guys thought it through to a speaker design .
I'm not sure if the first poster's use of "crossover" was a misnomer or not, based on the association with electronic music signals. Of course the spectral domain is very sensitive but still requires interpretation. I often presume when I see plots of distortion versus level that the rise as level decreases is simply the contribution of noise, but in some cases it may be crossover distortion. An actual spectrum at different levels of course conveys a lot more information.
That is an intriguing part, and if the minimum level for excitation of the inductor was a bit smaller I have an immediate application (and not for a displacement sensor).
The "not enough bass" complaint came up with a home theater system I played a part in. The careful limiter was completely readjusted by a guy who had worked for Velodyne, and made the system, for lack of a better word, sound far more aggressive. To each his own, but the end customer was satisfied. Sadly the product release was plagued with a very high initial failure rate on a purchased subsystem, and coupled with the convulsions in the economy and sudden restriction of easy credit, people were either buying bargain-basement soundbar/sub systems or had nearly-unlimited funds for high-end ones. The middle ground sales went to virtually zero. There was a desperate attempt to sell the motorized screen with its built-in LCR speakers as a standalone product, but the parent company just wanted the whole bad experience to go away. Oh well, it paid the bills for a while.
I have seen the first demo. It had too many holes in it and they have not reappeared with more in the intervening years. You can try this Audio DiffMaker which does the same essentially and see for yourself how challenging the tasks are. And how questionable the demo was.
The NXT think was also more of a myth than real. There were two other companies ahead of NXT at the time and NXT wound up buying both for serious money to keep the facade going. It was more of an investor Ponzi than real in the early period. They blew through some serious money. Now its the BMR in the Naim speaker. Interesting driver with good dispersion but not very good inner resolution (in my experience).
Some of the most accurate bass I've heard was at Toole's, who has an Infinity multichannel system including a number of subwoofers, and with the positioning carefully determined. He played a CD with the beginning of Also Sprach Zarathrustra, with the organ frankly recorded at way too high a level for demonstration purposes (I've heard a performance at Disney Hall that was in proper balance). But my instinctive reaction was to migrate toward a door jamb, despite a low-frequency cutoff that indicated that it was not in fact an earthquake. The only other time I've heard something comparable was at Lick Observatory early in the morning, when there was a relatively nearby static rocket engine test. Despite having only been briefly asleep after a night of observing, it definitely got everyone up for a while.
If the bass cannot rock your liver when required, what's the point?
The whole trick is to make it capable of big sound pressure levels AND stay as clean as possible. Big SPL is no problem, many do it for low prices, but what you get as bass is enough to make you cry more often than not.
A few days ago, I had the pleasure of hearing the Yamaha NS-1000 monitor again, after something like 35 years later on. The friend I was boogin' with complained by saying: "Where's the bass gone?". Initially, that might be the conclusion of many, I mean, one does expect a 12" woofer to be heard, right?
Well, it is, but the amount of overshoot and overhang is incredibly small, so what you do get is in fact uncommonly clean bass, which might lead you to believe there isn't enough of it.
The other problem is that this speaker is a know high difficulty level to drive, remember that it was one of two (with AR3a Improved) Otala used as examples of hard to drive speakers in his published works. Initially, the NS-1000 was driven by two Denon monoblocks, each rated at 150/220W into 8/4 Ohms, and much more on impulses into 2 Ohms. However, once we swapped it for a H/K Citation 24, with a nominally lower rating by about 2 dB, the NS-1000 took off and gave a very convincing bass performance. A typical case when specs tell you diddly. To be fair, I did beef the Citation a bit, when changing caps I replaced its 4 6,800uF caps with brand new 10,000 uF caps, upping the ante by a nice +47%.
All this might well apply to servoed bass speakers. If you rid them of excessive overhang, you might find they produce a much cleaner sounding bass, which however at first hearing might appear to be less bass. Makes you wonder what were you listening to when it had its initial bass.
The whole trick is to make it capable of big sound pressure levels AND stay as clean as possible. Big SPL is no problem, many do it for low prices, but what you get as bass is enough to make you cry more often than not.
A few days ago, I had the pleasure of hearing the Yamaha NS-1000 monitor again, after something like 35 years later on. The friend I was boogin' with complained by saying: "Where's the bass gone?". Initially, that might be the conclusion of many, I mean, one does expect a 12" woofer to be heard, right?
Well, it is, but the amount of overshoot and overhang is incredibly small, so what you do get is in fact uncommonly clean bass, which might lead you to believe there isn't enough of it.
The other problem is that this speaker is a know high difficulty level to drive, remember that it was one of two (with AR3a Improved) Otala used as examples of hard to drive speakers in his published works. Initially, the NS-1000 was driven by two Denon monoblocks, each rated at 150/220W into 8/4 Ohms, and much more on impulses into 2 Ohms. However, once we swapped it for a H/K Citation 24, with a nominally lower rating by about 2 dB, the NS-1000 took off and gave a very convincing bass performance. A typical case when specs tell you diddly. To be fair, I did beef the Citation a bit, when changing caps I replaced its 4 6,800uF caps with brand new 10,000 uF caps, upping the ante by a nice +47%.
All this might well apply to servoed bass speakers. If you rid them of excessive overhang, you might find they produce a much cleaner sounding bass, which however at first hearing might appear to be less bass. Makes you wonder what were you listening to when it had its initial bass.
Please define "overhang". We really all need to talk with understanding or its all lost. I understand it to mean overshoot and associated harmonic generation. Servos do not automatically create "overhang". In fact you need to mistune them significantly to get "overhang". That's really an implementation or knowledge problem on the part of the designer, not inherent in the technology.
You defined it yourself exactly as I meant it.
I understand servoes to do the opposite, to rid us of unwanted overhang, not to create it.
Also, I never implied it being inherent to any technology. However, that said, I must say I find AR's "acoustic suspension" to be particularly free of unwanted bass driver sallying after a transient, resulting in very clean and articulate bass.
Which is why I have owned AR speakers since 1972 to this day.
- Status
- Not open for further replies.