For a sine wave peak value and RMS value tell you precisely the same information. For other waveforms they do not. However, the whole point of FFT is to tell you what sine waves are present in your signal (amplitude and phase). There is no other information in that signal. How you interpret the information is up to you, but FFT does not destroy information. It may be that it doesn't present it in a way which you find useful, but that is a separate issue. By considering the waveform and the spectrum (both of which fully describe the signal, but in different ways) we can see most of the things of interest to us. However, you can regard these two as the two extremes of view (fully time, fully frequency) so between them are various wavelet models.RNMarsh said:
I hate statistical mechanics, but that is simply because I never could learn statistics. I have no doubt that it gives an accurate description of gases etc. There is a slight fudge in it: it is usually assumed that the ensemble average and time average are equal (ergodic theorem IIRC?). Fortunately this seems to be true for all physical systems we are interested in. Einstein was talking about quantum mechanics, not statistical mechanics? These are very different: SM says we don't know what each molecule is doing but we don't need to know while QM says we can't know because 'doing' is not definable in classical terms.Wavebourn said:
kT/2 per degree of freedom? I agree; degrees of freedom seem to be pulled like rabbits out of a hat until the numbers match. One of the earliest exposures to this I had was the calculation for Johnson noise in a resistor; it uses a terminated transmission line which then can disappear at the end of the calculation. This always looked like sleight of hand.scott wurcer said:
Curious me, thought of following him to the John Harrington and beat the C out of him ?
(not that I would do that to Lou Reed)
In classical statistics the basic principle is that as Laplace stated, "statistics is a thing we have due to lack of other knowledge", this being a belief in absolute determinism.
In statistical mechanics for instance the assumption is that a gas is an assemblage of elastic spheres in random motion undergoing Newtonian collisions, when you start calculating more than a few of these collisions the task rapidly gets out of hand and represents an impossible feat of calculation.
What statistical mechanics does is to point out that if you take a sufficient number of these spheres, you can describe what is happening as wave phenomena in a continuous smooth medium, the actual process is an absolutely determinable Newtonian one, its just that trying to calculate it is virtually impossible.
Historically degrees of freedom were also to explain observed phenomena.
The calculation of the speed of sound in air was wrong if you assumed that it consisted of elastic spheres, chemists however had discovered that the gases in the atmosphere were diatomic.
I think it was Boltzmann who then pointed out that if you then assumed two elastic spheres joined by a rigid rod, and then you assumed that each degree of freedom of such a system received equal amounts of energy then you could by this means accurately calculate the speed of sound in a mixture of gases.
Einstein was of course talking about quantum mechanics.
Heisenburg had shown that the basic unit that described the motion of particles, action, was described by a pair of non commutating matrices, that is the way around you multiplied them depended upon the answer you got,
Heisenberg when taking his results to his supervising professor is said to have remarked that he hoped he wasn't right, as far as anybody knows he was and still is.
rcw
In statistical mechanics for instance the assumption is that a gas is an assemblage of elastic spheres in random motion undergoing Newtonian collisions, when you start calculating more than a few of these collisions the task rapidly gets out of hand and represents an impossible feat of calculation.
What statistical mechanics does is to point out that if you take a sufficient number of these spheres, you can describe what is happening as wave phenomena in a continuous smooth medium, the actual process is an absolutely determinable Newtonian one, its just that trying to calculate it is virtually impossible.
Historically degrees of freedom were also to explain observed phenomena.
The calculation of the speed of sound in air was wrong if you assumed that it consisted of elastic spheres, chemists however had discovered that the gases in the atmosphere were diatomic.
I think it was Boltzmann who then pointed out that if you then assumed two elastic spheres joined by a rigid rod, and then you assumed that each degree of freedom of such a system received equal amounts of energy then you could by this means accurately calculate the speed of sound in a mixture of gases.
Einstein was of course talking about quantum mechanics.
Heisenburg had shown that the basic unit that described the motion of particles, action, was described by a pair of non commutating matrices, that is the way around you multiplied them depended upon the answer you got,
Heisenberg when taking his results to his supervising professor is said to have remarked that he hoped he wasn't right, as far as anybody knows he was and still is.
rcw
I met Mick Ronson in the bathroom that night, he was as drunk as Lou, but not as obnoxious, he told me to ignore Lou who was well known as a " Billy no mates" on occasion.
The big MBLs do have oversized imstruments at times, wonderfull detail, sound very smooth and as John says reproduce female voices very realistically. They are also very, very inefficent - their shown with big amps for a reason. At $250k for the speakers and another $250k for the matching MBL amplifiers (you do want them to match to stand a chance of getting them past " she who must be obeyed"). This is a speaker which is entirely OK to cash in your 401k for and enjoy them that evening, because when she finds out what you spent on them you'll never live long enough to retire. Oddly no expensive audio product owner's manual lists a set of possible excuses you need for your recent purchase or lists dress and jewellery shops in your area you may need in a hurry.
The big MBLs do have oversized imstruments at times, wonderfull detail, sound very smooth and as John says reproduce female voices very realistically. They are also very, very inefficent - their shown with big amps for a reason. At $250k for the speakers and another $250k for the matching MBL amplifiers (you do want them to match to stand a chance of getting them past " she who must be obeyed"). This is a speaker which is entirely OK to cash in your 401k for and enjoy them that evening, because when she finds out what you spent on them you'll never live long enough to retire. Oddly no expensive audio product owner's manual lists a set of possible excuses you need for your recent purchase or lists dress and jewellery shops in your area you may need in a hurry.
Last edited:
That's like complaining that a hammer is no good for interpreting Shakespeare.
What those "marvels" will do is tell you if your reproduction equipment is delivering the same signal to your loudspeakers as was encoded on your CD or if your speakers are spraying stuff around that shouldn't be there and obscuring the effect engineered into the recording. That's their function and they do it well, assuming the user is properly acquainted with the tools (much like a hammer, I suppose).
Ticknpop, I am always surprised that they sound so good, yet look so strange. However, some of the diagrams put up here did provide at least a glimpse of how they work.
Now, let me say something about amplifiers. As you probably already know, I first attempted to seriously design an extraordinary amplifier, in 1977 or 45 years ago.
I started at this time, because I worked for Ampex, and the complementary parts were available to me, for the first time.
Within a year, I had improved the design to look very similar to what many people make today, even on this website, so 44 years ago, I had a working Comp-Diff input stage driving a complementary push pull output stage. I used this little amp for several years, replacing my 'stock' Dyna Mk3. However, when I compared this little amp to a competing triode amp made by Radiocraftsman in the early 1950's, the Radiocraftsman sounded better to both my associates and me. This was a setback, BECAUSE I did NOT know why. I compared both amps for bandwidth, damping factor, stability, SMPTE IM distortion with increasing level, and the order of the IM byproducts. They were essentially similar, being virtually unmeasurable at 1W and lower (less than .005% IM and 104 SPL out of the speaker) slowly and monotonically increasing to perhaps 0.1% at 10W into an 8 ohm load.
Now WHY was there a difference in sound quality, and why didn't my direct coupled complementary design win the listening test, as there was a difference between the amps? At this time, perhaps 1969, I concluded that the ONLY significant difference between the two amps was that I used 40 dB of negative feedback, and the triode amp used about 20dB. This, of course meant that the triode amplifier itself was more intrinsically linear than the transistor amp even though I operated the output devices at 0.5A which was also the peak beta for the devices used. This meant that for the first 110 SPL or so, my amp ran class A. How could I ever have gotten louder than that in the listening room? We had neighbors, you know. In any case, I wanted to solve this dilemma.
About 1970, a paper was published in an IEEE Journal, by Dr.M.Otala that described that wide open loop bandwidth was necessary, and perhaps only 20 dB of feedback should be used in a power amp to reduce distortion. I thought to myself, maybe that is a good idea, since the Radiocraftsman amp had only 20 dB of feedback, maybe someday I should try to do this. Years went by, and while I made a number of power amp prototypes, none ever got built, commercially. I just had not gotten my protection circuits good enough to both sound OK and NOT let the amp break.
Then in about 1973, Matti Otala, and Jan Lohstroh (sp) wrote a paper that was published in the JAES that described an amplifier that appeared to have an open loop bandwidth of 20KHz, and apparently worked pretty well. Within the year, I had a chance to make my own small power amplifier as the tweeter driver for a 3 way all horn loaded speaker, so I decided to try for a good open loop bandwidth, 20 dB of overall feedback, and to use complementary jfets instead of bipolar transistors in the complementary differential input stage. The basic circuit ran about 1A of quiescent current, but was voltage swing limited to be about 20W in power, max. into 8 ohms. Since the tweeter we were using had something like 109 SPL/W, we didn't need much to drive it.
Well, this execution of a power amp (later to become the JC-3) wound up with a slew rate of 100V/us, maybe 5 times my earlier efforts with bipolar input devices. However, it had TOO MUCH FEEDBACK to fit in with what Matti recommended, so I had a choice: I could throw away the gain by putting a relatively low value resistor to ground from the drive stage, like Matti Otala and Ayre (today)do, OR maybe I could just feedback the excess gain and linearize the input stage further, as well as VOLTAGE DRIVE the Complementary Darlington output stage, which I knew from experience would be more linear than CURRENT driving it, because then you really don't have a follower, but you have a BETA MULTIPLIER that is extremely sensitive to changes in BETA over the current range.
So I added the resistor (the one that is being disputed here) so that my overall feedback would be 20dB, and (at least I thought) that the open loop bandwidth would be 20K or so. I could have been mistaken with this assumption, and maybe I should have just thrown the extra gain away in order to 'do it right' but the results were pretty good, in any case, compared to other amps that we had in the lab like an SAE and a couple of Marantz amps. (more to come)
Now, let me say something about amplifiers. As you probably already know, I first attempted to seriously design an extraordinary amplifier, in 1977 or 45 years ago.
I started at this time, because I worked for Ampex, and the complementary parts were available to me, for the first time.
Within a year, I had improved the design to look very similar to what many people make today, even on this website, so 44 years ago, I had a working Comp-Diff input stage driving a complementary push pull output stage. I used this little amp for several years, replacing my 'stock' Dyna Mk3. However, when I compared this little amp to a competing triode amp made by Radiocraftsman in the early 1950's, the Radiocraftsman sounded better to both my associates and me. This was a setback, BECAUSE I did NOT know why. I compared both amps for bandwidth, damping factor, stability, SMPTE IM distortion with increasing level, and the order of the IM byproducts. They were essentially similar, being virtually unmeasurable at 1W and lower (less than .005% IM and 104 SPL out of the speaker) slowly and monotonically increasing to perhaps 0.1% at 10W into an 8 ohm load.
Now WHY was there a difference in sound quality, and why didn't my direct coupled complementary design win the listening test, as there was a difference between the amps? At this time, perhaps 1969, I concluded that the ONLY significant difference between the two amps was that I used 40 dB of negative feedback, and the triode amp used about 20dB. This, of course meant that the triode amplifier itself was more intrinsically linear than the transistor amp even though I operated the output devices at 0.5A which was also the peak beta for the devices used. This meant that for the first 110 SPL or so, my amp ran class A. How could I ever have gotten louder than that in the listening room? We had neighbors, you know. In any case, I wanted to solve this dilemma.
About 1970, a paper was published in an IEEE Journal, by Dr.M.Otala that described that wide open loop bandwidth was necessary, and perhaps only 20 dB of feedback should be used in a power amp to reduce distortion. I thought to myself, maybe that is a good idea, since the Radiocraftsman amp had only 20 dB of feedback, maybe someday I should try to do this. Years went by, and while I made a number of power amp prototypes, none ever got built, commercially. I just had not gotten my protection circuits good enough to both sound OK and NOT let the amp break.
Then in about 1973, Matti Otala, and Jan Lohstroh (sp) wrote a paper that was published in the JAES that described an amplifier that appeared to have an open loop bandwidth of 20KHz, and apparently worked pretty well. Within the year, I had a chance to make my own small power amplifier as the tweeter driver for a 3 way all horn loaded speaker, so I decided to try for a good open loop bandwidth, 20 dB of overall feedback, and to use complementary jfets instead of bipolar transistors in the complementary differential input stage. The basic circuit ran about 1A of quiescent current, but was voltage swing limited to be about 20W in power, max. into 8 ohms. Since the tweeter we were using had something like 109 SPL/W, we didn't need much to drive it.
Well, this execution of a power amp (later to become the JC-3) wound up with a slew rate of 100V/us, maybe 5 times my earlier efforts with bipolar input devices. However, it had TOO MUCH FEEDBACK to fit in with what Matti recommended, so I had a choice: I could throw away the gain by putting a relatively low value resistor to ground from the drive stage, like Matti Otala and Ayre (today)do, OR maybe I could just feedback the excess gain and linearize the input stage further, as well as VOLTAGE DRIVE the Complementary Darlington output stage, which I knew from experience would be more linear than CURRENT driving it, because then you really don't have a follower, but you have a BETA MULTIPLIER that is extremely sensitive to changes in BETA over the current range.
So I added the resistor (the one that is being disputed here) so that my overall feedback would be 20dB, and (at least I thought) that the open loop bandwidth would be 20K or so. I could have been mistaken with this assumption, and maybe I should have just thrown the extra gain away in order to 'do it right' but the results were pretty good, in any case, compared to other amps that we had in the lab like an SAE and a couple of Marantz amps. (more to come)
Now, about my experience with the Otala power amplifier.
In 1975, the institute I worked at was closing down, and our lab closed with it. I went to work for someone else in Switzerland, so I stayed in the area, and there was an AES conference in Zurich the next year.
Because they had heard of me, somehow, two young guys from Norway visited me with an amp prototype that they wanted me to addition. At the time I had tried a number of the 'better' commercial transistor amps and none sounded good with my STAX Electrostatic headphones. They all sounded kind of 'phony' and gave me listening fatigue, WHEN I used the headphones. Through the Magnepans they were sort of OK.
Then these young guys let me try the Otala amp which they had copied from the AES preprint, and let me listen to it. It was a revelation! It sounded really good with my electrostatic headphones too! It wasn't very powerful, but it sounded 'right' at normal listening levels. I just had to have this amp.
They did NOT want to sell it to me, because this was their prototype, and they had to take it back across the border, where they registered it, apparently.
Well, I was very convincing, and I offered them real money AND a Quatre amp chassis that they could use for prototyping, and get them across the border. This amp was built in a very amateur way. It is amazing that it stayed working at all, BUT it did sound good, AND with a few substitutions along the way, I kept it until the firestorm in 1991 took it, my JC-80, my WATT-Puppies, and STAX electrostatic headphones, etc.
Yes, this little Electrocompaniet sounded better than just about anything that I could afford, for more than 15 years, and it was even good enough to drive a set of WATT's with remarkable sound quality.
Now what did Jan and Matti do RIGHT, to make me so enthusiastic? I knew of Otala, but I don't think I ever met the guy, before buying the amp. I had tried some of his concepts both in the GD equipment, the Levinson JC-2, and in the Institute power amp with some success, but really this little amp kicked TAIL, sonically, and I will never be sorry for the purchase. About 20 years ago, I found another Otala amp, virtually the same as the one I lost, and I have it in my office system driving a pair of Sequerra speakers. Pound for pound, this little amp beats my HCA-1000 Parasound amp, which is rather embarrassing, but it is important to be honest about such things. (end of tale)
In 1975, the institute I worked at was closing down, and our lab closed with it. I went to work for someone else in Switzerland, so I stayed in the area, and there was an AES conference in Zurich the next year.
Because they had heard of me, somehow, two young guys from Norway visited me with an amp prototype that they wanted me to addition. At the time I had tried a number of the 'better' commercial transistor amps and none sounded good with my STAX Electrostatic headphones. They all sounded kind of 'phony' and gave me listening fatigue, WHEN I used the headphones. Through the Magnepans they were sort of OK.
Then these young guys let me try the Otala amp which they had copied from the AES preprint, and let me listen to it. It was a revelation! It sounded really good with my electrostatic headphones too! It wasn't very powerful, but it sounded 'right' at normal listening levels. I just had to have this amp.
They did NOT want to sell it to me, because this was their prototype, and they had to take it back across the border, where they registered it, apparently.
Well, I was very convincing, and I offered them real money AND a Quatre amp chassis that they could use for prototyping, and get them across the border. This amp was built in a very amateur way. It is amazing that it stayed working at all, BUT it did sound good, AND with a few substitutions along the way, I kept it until the firestorm in 1991 took it, my JC-80, my WATT-Puppies, and STAX electrostatic headphones, etc.
Yes, this little Electrocompaniet sounded better than just about anything that I could afford, for more than 15 years, and it was even good enough to drive a set of WATT's with remarkable sound quality.
Now what did Jan and Matti do RIGHT, to make me so enthusiastic? I knew of Otala, but I don't think I ever met the guy, before buying the amp. I had tried some of his concepts both in the GD equipment, the Levinson JC-2, and in the Institute power amp with some success, but really this little amp kicked TAIL, sonically, and I will never be sorry for the purchase. About 20 years ago, I found another Otala amp, virtually the same as the one I lost, and I have it in my office system driving a pair of Sequerra speakers. Pound for pound, this little amp beats my HCA-1000 Parasound amp, which is rather embarrassing, but it is important to be honest about such things. (end of tale)
I remember some designs for low t.i.m. power amplifiers.
A popular one here in Oz slowed down the long tailed pair by putting a rc circuit between the collectors.
Electronics types I knew said that the principle was quite simple, the major trouble was that the amplifier had a open loop rise time that was severely curtailed by the 3055 and 2955 transistors of the time.
The tendency toward t.i.m. could be fixed by slowing down the input so that the feedback for fast rising signals arrived about the same time as the input signal made its way to the point where feedback was applied, also making the open loop zero crossing frequency as high as you could get it in the driver and output stages maximised the amount of feedback you could get at those frequencies.
Whether it worked or the whole thing had any validity anyway is a thing that is still debated, but equivalent modern bipolar devices have an ft in the 35MHz. region, the open loop response then very definitely is due to the driver stage.
rcw
A popular one here in Oz slowed down the long tailed pair by putting a rc circuit between the collectors.
Electronics types I knew said that the principle was quite simple, the major trouble was that the amplifier had a open loop rise time that was severely curtailed by the 3055 and 2955 transistors of the time.
The tendency toward t.i.m. could be fixed by slowing down the input so that the feedback for fast rising signals arrived about the same time as the input signal made its way to the point where feedback was applied, also making the open loop zero crossing frequency as high as you could get it in the driver and output stages maximised the amount of feedback you could get at those frequencies.
Whether it worked or the whole thing had any validity anyway is a thing that is still debated, but equivalent modern bipolar devices have an ft in the 35MHz. region, the open loop response then very definitely is due to the driver stage.
rcw
That is way off the point, if you invent a meter that judges an interpretation of say Mahler's 2nd you could probably find a buyer.
At times.
I dunno. That's often the case, but my Altec A5s are pretty big - similar to the MBL - and don't give me that over-sized sound. And I do sit too close. That's what I found striking about the MBL, it's extreme resolution, detail, realistic harmonic structure, etc. And yet it sounds too big. Why? A set-up problem or something inherent in the technology?
Most loudspeakers are anything but omnidirectional. When one looks at the problem of sound reproduction (as opposed to simulation of a musical instrument with a complex pattern of directivity) omni is not necessarily better, unless somehow the recording has been made to exploit this.
Note the comments made about solo "instruments", probably indicative of what would be the best candidates for exploiting the MBLs.
I don't see big difference.
But there is. One is, "It's too difficult to find out as a practical matter but you don't need to," the other is, "The question itself, while grammatically correct, has no physical meaning." The latter is a tough pill for many to swallow, but that's the universe we're living in.
Fred Hoyle had an interesting take on QM which he was able to expouse when he was the guest essayist in a volume of the Annual Reviews of Astronomy and Astrophysics. These contributions are invited and not refereed. Most astronomers I knew were embarrassed by the piece; I found it intriguing, if not quite as much of an epiphany as did Sir Fred.
EDIT: The piece is in the 1982 ARAA and titled The Universe: Past and Present Reflections. He also plugs JBL at one point, mentioning hearing some speakers they had at Kitt Peak.
EDIT: The piece is in the 1982 ARAA and titled The Universe: Past and Present Reflections. He also plugs JBL at one point, mentioning hearing some speakers they had at Kitt Peak.
Last edited:
Radials sound 'big' because they give the perspective
of being amongst the orquestra rather than in the
audience.
How long is the song? If you accept a 5 min CD version as being acceptible, you have 52.92M-byte as a acceptible limit on the total information. The 13.23M-sample FFT of each channel contains the same information. Maybe you can't use this frequency domain information for anything useful, that is not the point.
Maybe we should talk about IIR filters, four numbers streaming through a file with simple multiply and adds can apply an RIAA characteristic to better than .01dB and fractional degrees of phase. Seems to good to be true.
It was an example of usage of one of tools for what it can be used. However, screwdriver can be used as a rudimentary hammer, but it is not optimal.
Mostly --- its the omni characteristic and the larger speaker driver radiating area spread out (vertically).
Does not matter.
Look how it is "simple" and "elegant":
https://docs.google.com/viewer?a=v&...&sig=AHIEtbRztRsfQp-cwg_8Bqk_lU9RNrqtLQ&pli=1
Actually, it can be useful. Here's one example: say you're worried about the phase shift from the anti-aliasing and anti-imaging filters. Multiply the frequency domain data by the inverse phase function, then back-transform to the time domain. Voila! Flat phase curve!
- Status
- Not open for further replies.