Hi,
And I made it clear that there is already tons written on that and based on the Index the Book proposed will not advance the state of the art over what was extant 30 Years ago.
So I am suggesting to can the project and look forward, not back.
I once paid my own good money for a book on amplifier design which not only had less coverage than the Wireless World articles which I had already paid for and owned, worse, all the really interesting (at least to the lunatic fringe I count myself among) bits from Wireless World where missing.
The book covered in excruiating detail minor improvements to a (then) 50 Years old topology of power amplifer, which had been rendered obsolete in performance terms at least 20 years earlier, outside the arena of "corporate interests that only want to push garbage on the "sheeple" , who know not what good Hi-Fi sounds like! ".
Having found this, the book remained after the first time entirely unopened on my shelf for the better part of a decade, until it was thrown into the garbage, having been opened exactly once, to be skimp-read.
Here I see another such work coming on, with as much relevance (at least I will not buy it this time).
However I feel I rather have my say now to give the author time to reconsider (He did not offer such an opportunity last time around), to make sure the book remains relevant for more than 10 minutes after opening it. And I would honestly prefer he would apply his considerable and much appreciated talents in research and writing to a cause that looks forward and advances the state of the art, instead of offering a compendium of obsolete knowledge, that merely restates what was well known and understood decades ago!
Some may consider this Lèse majesté (after all, we are talking here about Him, HinSelf), but to me authority accrues not from position or background, but from actual demonstration in their field.
As a direct personal message to DS, I do respect your work, it is scientific and meticulous in ways I can never be bothered with and I have learned much from your articles in WW. You can still be at the cutting edge, just look forward, not back. Past that, if you can spare the time, I'll buy the first three rounds at Murphys in Tsim Sha Tsui if you can make the short trip, let me know.
Ciao T
And I made it clear that there is already tons written on that and based on the Index the Book proposed will not advance the state of the art over what was extant 30 Years ago.
So I am suggesting to can the project and look forward, not back.
I once paid my own good money for a book on amplifier design which not only had less coverage than the Wireless World articles which I had already paid for and owned, worse, all the really interesting (at least to the lunatic fringe I count myself among) bits from Wireless World where missing.
The book covered in excruiating detail minor improvements to a (then) 50 Years old topology of power amplifer, which had been rendered obsolete in performance terms at least 20 years earlier, outside the arena of "corporate interests that only want to push garbage on the "sheeple" , who know not what good Hi-Fi sounds like! ".
Having found this, the book remained after the first time entirely unopened on my shelf for the better part of a decade, until it was thrown into the garbage, having been opened exactly once, to be skimp-read.
Here I see another such work coming on, with as much relevance (at least I will not buy it this time).
However I feel I rather have my say now to give the author time to reconsider (He did not offer such an opportunity last time around), to make sure the book remains relevant for more than 10 minutes after opening it. And I would honestly prefer he would apply his considerable and much appreciated talents in research and writing to a cause that looks forward and advances the state of the art, instead of offering a compendium of obsolete knowledge, that merely restates what was well known and understood decades ago!
Some may consider this Lèse majesté (after all, we are talking here about Him, HinSelf), but to me authority accrues not from position or background, but from actual demonstration in their field.
As a direct personal message to DS, I do respect your work, it is scientific and meticulous in ways I can never be bothered with and I have learned much from your articles in WW. You can still be at the cutting edge, just look forward, not back. Past that, if you can spare the time, I'll buy the first three rounds at Murphys in Tsim Sha Tsui if you can make the short trip, let me know.
Ciao T
Hi,
Tell that to DS, his employer (last I knew) was Cambridge Audio, aka Audio Partnership, aka. MIC.
To my ears the best state of the art still fails to match 1950's tube stuff, but that is just me.
All I know is that if you combine current understandings of the acoustic side (which have not really advanced since I learned it in the 80's, but which have become vastly more accessible) with the latest digital tech we can literally transcend the boundaries that have limited the reproduction side so far.
Of course, the value of such transcendence in the face of current widespread recording practice may be questionable, but that is a whole other debate.
All is analogue (the signal), all is digital, that is the mechanisms employed by nature/evolution/G*d/random chance (pick one/any/all to your taste) in the original, good oldfashioned MK I human ear.
The debate of digital, analogue etc. is basically a big misunderstanding. The debate about quality matters greatly, to some, but is independent of technology as such.
As someone quite familiar with both amplifers that are claimed to be blameless, amplifiers that quite clearly are very blame-able and indeed digital amplifiers (as in fully digital amplifiers, not analogue input switching amp's) that may or may not be blameable, I know which in subjective terms comes last.
It is the Amplifier I would rarely if ever blame for producing good sound. Which of the three that is - well, you listen and you decide for yourself.
Ciao T
Tell that to DS, his employer (last I knew) was Cambridge Audio, aka Audio Partnership, aka. MIC.
To my ears the best state of the art still fails to match 1950's tube stuff, but that is just me.
All I know is that if you combine current understandings of the acoustic side (which have not really advanced since I learned it in the 80's, but which have become vastly more accessible) with the latest digital tech we can literally transcend the boundaries that have limited the reproduction side so far.
Of course, the value of such transcendence in the face of current widespread recording practice may be questionable, but that is a whole other debate.
All is analogue (the signal), all is digital, that is the mechanisms employed by nature/evolution/G*d/random chance (pick one/any/all to your taste) in the original, good oldfashioned MK I human ear.
The debate of digital, analogue etc. is basically a big misunderstanding. The debate about quality matters greatly, to some, but is independent of technology as such.
As someone quite familiar with both amplifers that are claimed to be blameless, amplifiers that quite clearly are very blame-able and indeed digital amplifiers (as in fully digital amplifiers, not analogue input switching amp's) that may or may not be blameable, I know which in subjective terms comes last.
It is the Amplifier I would rarely if ever blame for producing good sound. Which of the three that is - well, you listen and you decide for yourself.
Ciao T
Yes. I have only just been made aware of the Amazon announcement. A similar announcement has appeared at Focal Press:
Focal Press: The Design of Active Crossovers - Book
Looks like I'd better speed up the writing.
Not only did I have no control over it, but I was unaware that that compilation had been produced until I saw it in a bookshop. I wholly agree it is a train-wreck of a book, with an idiotic title, and I told my publishers in no uncertain terms what I thought of it. Remarkably, some of the Amazon reviewers seem to have liked it.
I think you might find this compilation rather better:
Audio Engineering Explained: Amazon.co.uk: Douglas Self: Books
When a compilation is put together, you naturally use sample chapters from the books unchanged. Since you are hoping people will buy the books it would not be honest to do anything else.
Last edited:
I also agree that building an analog active crossover is not likely to be a project anymore. But building a digital one is definitely also not your everyday project, as you need professional circuit board layout. How many powerful DSPs have you seen in DIP-format?.
But why build one from scratch when you can buy one much cheaper (unless you are an OEM) and invest all that project time gained in setting it up correctly?
So please include as much information as possible on how to set up a crossover (digital or analog) correctly, because I think it is this information that many DIYers needs nowadays…
But why build one from scratch when you can buy one much cheaper (unless you are an OEM) and invest all that project time gained in setting it up correctly?
So please include as much information as possible on how to set up a crossover (digital or analog) correctly, because I think it is this information that many DIYers needs nowadays…
Hello Douglas.
I'm glad that new interesting book will be available soon.
I was really impressed that you paid big attention to cover noise related topics of filter design. This seems to be very important but underestimated topic.
It is hard to tell what is exactly covered just by looking on TOC, but i think that "Loudspeaker equalization" section is really "Loudspeaker and listening room equalization".
I could not find noise related topics in equalization section. Some equalizer circuits are very prone to noise and sum of 30 bands may create big problems, until big attention is payed to level diagram, design of filters and sum circuit (something very similar to mixing consoles).
Also, I can recall some works in the 80's, trying to satisfy phase, SPL and power requirements for crossover networks. Some optimization was suggested to achieve possible compromises. I think that Robert Bullock III wrote a couple of interesting papers. I don't have assess to his papers right now, but probably Loudspeaker-Crossover Systems: An Optimal Crossover Choice is one of them. Do you plan to include this kind of crossovers in your book?
I'm glad that new interesting book will be available soon.
I was really impressed that you paid big attention to cover noise related topics of filter design. This seems to be very important but underestimated topic.
It is hard to tell what is exactly covered just by looking on TOC, but i think that "Loudspeaker equalization" section is really "Loudspeaker and listening room equalization".
I could not find noise related topics in equalization section. Some equalizer circuits are very prone to noise and sum of 30 bands may create big problems, until big attention is payed to level diagram, design of filters and sum circuit (something very similar to mixing consoles).
Also, I can recall some works in the 80's, trying to satisfy phase, SPL and power requirements for crossover networks. Some optimization was suggested to achieve possible compromises. I think that Robert Bullock III wrote a couple of interesting papers. I don't have assess to his papers right now, but probably Loudspeaker-Crossover Systems: An Optimal Crossover Choice is one of them. Do you plan to include this kind of crossovers in your book?
Ok, it's under equalizers. I guess I missed it when focusing on the "how loudspeakers work" chapter. I apologize.
The point I was trying to make is that Rauch filters (IMHO a more appropriate name, as you could call any crazy filter MFB) can be inconvenient for non-integrated implementations, compared to Sallen-Key equivalents. I personally dislike PCB design with 0.5pF capacitors defining filter transfer curves, whereas on an IC this is no problem at all. This fact is often omitted (and not very obvious until you actually start calculating component values).
Awww.... :-(
I hope you still include "advanced reading" references for such topics as GICs.
Ok, please allow me to use the Wilson Duette as an example. It has a passive XO, outside the main cabinet and mechanically fixed inside a solid block of epoxy. Still, mr. Wilson decided that the customer should be able to configure the XO according to how close the unit is to a wall (as mr. Wilson is aware, many customers are to some degree unhappily married and therefore constrained in their loudspeaker placement choices).
For this purpose, the product is delivered with a bunch of different resistors that the customer can insert in the extra set of terminals that the XO box has. Effective, but hardly elegant in an acoustic sense. What the high-end customer would really like is not just a variable tweeter amplitude, but with that some predefined changes in XO frequency and phase difference between the two drivers. Implementing this passively gets complicated and expensive quite quickly and will probably prohibit the use of the epoxy constructions around all of the reactive components.
In an active crossover, we could circumvent such issues. One way would be to implement all configurability with dirt-cheap voltage dividers in either jumpers or microcontrollers (costing approx. US$1.50, oscillator included). The resulting voltages could then be (filtered and) fed to VCFs. The nice thing about this is that we don't need large numbers of op-amps for each setting or linear configurable passives (especially varactors would be painful) in the signal path.
I imagine this is the approach Genelec could have taken for its 8040A and 8050A. They calibrate the XO together with the drivers in the factory anyway. This is cheaper than requiring strict tolerances on both drivers and (signal path) filter component values.
In addition, one could imagine that one would, in particular situations, prefer the XO to behave differently for sharp transients. This is, for example, the case in digital filters inside the emmLabs CD players and DACs. However, the technique is not limited to digital filters. With an amplitude/attack detector and a VCF the same thing can be done in analog (active only, passive cannot do this within reasonable constraints). The fact that nobody does this does not mean it's a bad idea. One can argue that in the end we've built a synthesizer and not an XO, but I can imagine this being a substantial extra degree of design freedom over a passive XO and therefore worth mentioning.
When I say LC-filter prototype I mean the LC-filter with inductors replaced by GIC or gm-C equivalents. The main advantage of LC-filters over RC-filters is sensitivity to variations in component values, and this property carries over to LC-prototypes with the inductors replaced by active equivalents. So, considering calibration time, LC-filter prototypes are not necessarily an expensive way of getting things done.
Concerning nonplanar circuit topologies, I will use an LC time-delay filter as an example (I will leave out ALL of the algebra!). Consider two input terminals (IN+ and IN-) and two output terminals (OUT+ and OUT-). We connect a loudpeaker driver, impedance assumed real, to the output terminals and a signal voltage source with unspecified series resistance at the input terminals. Next, we define the actual filter. Between IN+ and OUT+ we connect a parallel LC network (parallel resonant tank). We do the same between IN- and OUT-. Then we connect a series LC network (series resonant) between IN+ and OUT-. We also connect a series LC network between IN- and OUT+. This is our filter.
Now, why is this a nonplanar circuit topology and why would anyone care? A planar topology is one that can be drawn on a sheet of paper such that all intersecting lines correspond to the same node, in other words, we may draw dots on all line intersections. This means that all circuit meshes enclose empty bits of paper. This allows straightforward mesh/node equations to algebraically derive expressions for things like impedances and transfer functions. For me personally, this also allows intuitive feeling of signal flows and topology corrections, in case they are necessary.
A nonplanar topology, on the other hand, necessarily contains one or more line intersections where the intersecting lines belong to different nodes (so we cannot draw dots on these intersections :-( ). This translates to tedious algebra and for me often breaks the intuitive feel for the circuit. The LC time-delay filter has one such intersection; depending on how you draw the circuit it's either between the input nodes or between the output nodes. Therefore, it is nonplanar and consequently presents significant algebraic complications.
Since it's so difficult, it's the one general thing I would want explained in a reference book on filters, at least with the algebra worked out for one example. If it's relevant for this particular book depends on how many people actually use or consider using active implementations of LC time-delay filter prototypes. I guess not too many.
--
Maybe this post could have been shorter; I hope it at least makes for a slightly entertaining read...
Doug, I would be really interested in what you have to comment on my earlier point #1, concerning acoustic flatness. Will there be (is there enough space for...) some rules-of-thumb for increased acoustic efficiency at the X-over point for some standard loudspeaker configurations?
--
Greetz,
MatchASM
Maybe I'm being dim but I don't see how capacitor values that small are going to crop up in audio filters.
Good point.
It could be done with VCAs but there would be noise & distortion compromises. Switched settings would be better.
I don't think many people would think that was a good idea.
Never seen one, as far as I recall. The chapter on delay filters is already looking like one of the longer ones, and I don't think I can fit this in.
Not sure I'm following you. Can you expand on that?
--
Greetz,
MatchASM[/QUOTE]
If none of your Rauch filter designs ever end up with uncomfortably small capacitor values, there is indeed no point in bringing up this issue. In TV filter designs I have seen it is an issue, and I imagine it would be the same for a high-order elliptic audio filter (if ever you would choose to use such a filter).
Really? The analog A&H filters I have heard sound neither noisy nor distorted to me. Perhaps they have a patented way of making them this good; I never bothered to check. If there is a patent out there, then you might be correct in saying there are unavoidable noise and distortion trade-offs for the common XO designer.
Ok, I understand that I'm a minority here; OTOH a short mention of a potential advantage of actives vs. passives wouldn't be too painful IMHO.
Surely there must be some sort of rule-of-thumb that says how many dB acoustic efficiency correction is to be expected per configuration, for such standard arrangements as:
-tweeter adjacent to a 7" midwoofer,
-tweeter adjacent to a 5.5" midwoofer,
-tweeter adjacent to two 5.5" midwoofers in MTM configuration,
-tweeter adjacent to two 3" midranges in MTM configuration,
In all of the above, tweeter must be categorized according to 25mm dome, ring radiator or ribbon.
I assume that your approach will always be to aim the main acoustic lobe directly at the listener (or at least under a constant offset angle). If not, then the number of cases becomes a bit large and giving rules-of-thumb becomes tedious (and my question meaningless).
--
Greetz,
MatchASM
Well, TV does use rather higher frequencies than audio. In the last crossover I designed the smallest capacitor doing actual filtering was 2n2.
I still don't think I'm getting this. Are you talking about the dB SPL/Volt of the drive units?
Most of the thread has gone over my head I must admit, but having a electronic XO with the ability to tailor the XO to different combinations of mid/tweeter I can see as a definite advantage, probably done as timing delays,
One set slope may not suit all applications, I am aware that this can be done digitally but I assume we are designing discrete circuits here.
my experience with XOs is limited to Behringer gear if i had the money I would like many more features than my units possess
One set slope may not suit all applications, I am aware that this can be done digitally but I assume we are designing discrete circuits here.
my experience with XOs is limited to Behringer gear if i had the money I would like many more features than my units possess
@DS
As some already have perfectly outlined, you seem to be committed to kind of an museal artwork - nothing to add there.
Maybe - being in the topic - you might just do some chapters for your next book at weekend - which hopefully is more usable and valuable for saw-dust-DIYers?
- copy and paste a chapter about classic XO filters and how they blend from your current work as a starter
- do another copy and paste chapter about issues of group delay / phase behaviour, transient behaviour and resulting lobing
- add a chapter about the philosophy of response shaping
- add a chapter about the limits in digital processing as for example filter length issues and when CMP behaviour enters the picture
- close with a summery / outlook of current available DSP and PC based systems.
Should not take you more than a couple of days, and might be way more "in time"...
Just my 2ct
Michael
As some already have perfectly outlined, you seem to be committed to kind of an museal artwork - nothing to add there.
Maybe - being in the topic - you might just do some chapters for your next book at weekend - which hopefully is more usable and valuable for saw-dust-DIYers?
- copy and paste a chapter about classic XO filters and how they blend from your current work as a starter
- do another copy and paste chapter about issues of group delay / phase behaviour, transient behaviour and resulting lobing
- add a chapter about the philosophy of response shaping
- add a chapter about the limits in digital processing as for example filter length issues and when CMP behaviour enters the picture
- close with a summery / outlook of current available DSP and PC based systems.
Should not take you more than a couple of days, and might be way more "in time"...
Just my 2ct
Michael
Thank you all for your interest and you comments.
I would like to say a word or two on the analogue/DSP issue. Firstly I have nothing whatsoever against the DSP approach. It was me that did the initial digital mixing console work at Soundcraft on the Motorola 56001 processor back in the 80's. Clearly some functions such as pure time-delay are easier in the digital world, and there is unquestionably the advantage of complete precision in setting filter characteristics with no worries about capacitor tolerances and so on.
I have therefore structured the book so that at least two-thirds of it is applicable to either analogue or DSP. You study the concepts, you choose a crossover type, and find that at some point you need, say, a second-order Butterworth filter. It is then your choice whether to break out the 5532s or start coding for a DSP. I don't think it would be appropriate to start getting into the details of implementation on different processors. Another point is that there are a lot of DSP engineers about, but high-end analogue expertise is, shall we say, somewhat harder to find. I therefore make no apology for including three or four chapters specific to the analogue approach.
I certainly do not accept that analogue circuitry is in any way obsolete. The book will contain many novel techniques for improving the performance of opamp circuitry, and I have attempted to bring these together in a demonstration crossover design in Chapter 19 at the end of the book. The measured results for a version made almost entirely out of 5532s (there are a couple of LM4562s inserted where they will do the most good) show a signal/noise ratio of 117.5 dB for the HF path output, -122.2 dB for the MID output, and -127.4 dB for the LF output, which I think shows that using all the above noise reduction techniques can give some pretty stunning results. I don't pretend to be bang-up-to-date on converter technology, but I suspect that such a performance would be hard to beat by DSP.
In addition, I find it very hard to get enthusiastic about digital power amplifiers. Their distortion performance is much inferior to that of a good Class-B power amplifier, and they generate unhappily high levels of RF noise.
A Happy New Year to you all.
I would like to say a word or two on the analogue/DSP issue. Firstly I have nothing whatsoever against the DSP approach. It was me that did the initial digital mixing console work at Soundcraft on the Motorola 56001 processor back in the 80's. Clearly some functions such as pure time-delay are easier in the digital world, and there is unquestionably the advantage of complete precision in setting filter characteristics with no worries about capacitor tolerances and so on.
I have therefore structured the book so that at least two-thirds of it is applicable to either analogue or DSP. You study the concepts, you choose a crossover type, and find that at some point you need, say, a second-order Butterworth filter. It is then your choice whether to break out the 5532s or start coding for a DSP. I don't think it would be appropriate to start getting into the details of implementation on different processors. Another point is that there are a lot of DSP engineers about, but high-end analogue expertise is, shall we say, somewhat harder to find. I therefore make no apology for including three or four chapters specific to the analogue approach.
I certainly do not accept that analogue circuitry is in any way obsolete. The book will contain many novel techniques for improving the performance of opamp circuitry, and I have attempted to bring these together in a demonstration crossover design in Chapter 19 at the end of the book. The measured results for a version made almost entirely out of 5532s (there are a couple of LM4562s inserted where they will do the most good) show a signal/noise ratio of 117.5 dB for the HF path output, -122.2 dB for the MID output, and -127.4 dB for the LF output, which I think shows that using all the above noise reduction techniques can give some pretty stunning results. I don't pretend to be bang-up-to-date on converter technology, but I suspect that such a performance would be hard to beat by DSP.
In addition, I find it very hard to get enthusiastic about digital power amplifiers. Their distortion performance is much inferior to that of a good Class-B power amplifier, and they generate unhappily high levels of RF noise.
A Happy New Year to you all.
Hi Douglas,
I could not think of any application where I would want that in speakers. No Speaker I ever designed was able to operate with textbook crossovers unless massive amounts of heavy handed EQ where added.
Textbook crossovers (active or passive) just do not work with real drivers.
And it is by far smarter to modify the filter to do as much of the EQ and response shaping required in combination with the filtering (especially if doing it analogue, be it passive or active, but also in DSP as it saves calculations or DSP horsepower) then to daisychain textbook filters and textbook equalisers 20 stages deep.
Rapidly heading there. Honest.
Well, using TI's all Digital Amplifiers at -24dB attenuation (which is pretty common for normal listening) beats this comprehensively, unless you place the volume control after your crossover, to preserve the SNR, in which case you can better the TI Amp's by a little bit. And I would not claim the TI solution to be state of the art (it is meant for multichannel AV Receivers).
Modern DAC Chips easily measure lower levels of noise than those that you quote, the key to preserving it is good design of analogue stages, if one insists on using analogue amplification.
Now this would be a worthy topic, especially in the light of the ton's of ultrasonic output modern converters output and the tendency of Op-Amp's to have rapidly rising distortion with rising frequency (which you very aptly illustrated in your WW articles), which has quite dramatic implications that normal analogue design works tend not to sufficiently address.
But to be honest, it really seems SO 20th Century to do it that way. If we really want to do old-tech, where are the tubes?
Some specific users with specific performance requirements or prejudices may disagree with the all digital approach (e.g. me, I also disagree with using Op-Amp's of course, or transistors unless absolutely unavoidable), but for mainstream use, even in extreme High End contexts the all digital approach makes tremendous sense (even on performance grounds).
And there are enough "analogue" circuit tricks in switching amplifiers and similar areas that could be gainfully documented (I still remember the little trick of yours to shift the signal take-off for the Miller Compensation Cap to the Amplifier Output which makes a whole lot of difference with not a single component added) in combination with the techniques for "non-textbook" crossovers (and how to synthesise them from the driver response) that would make a book on active speakers very relevant in the 21st century.
Actually, to beat that would be a little non-trivial, due to analogue stage issues that are parallel to those in your crossover (noise, mainly(, but not in any way unsolvable or even really difficult.
Throw together a PCM1794 and a single OPA1632 per output in a competent way and you have equalled if not exceeded the performance you quote at LF and trashed it at MF & HF, at very modest cost.
Well, arbitrarily low distortion does not guarantee good sound and relatively high levels of distortion (such as found in tube amplifiers) do not guarantee bad sound.
Given that arbitrarily low THD is trivial to achieve using modern solid state circuitry (as in "more advanced than the 1950's 'Lin' Amplifier") I find it hard to get excited about low THD either.
The subject of distortion and sound quality warrants more investigation, as in the end it is the percieved sound quality that matters.
I have noticed that some of the faster switching, lower power (meaning they run very close to 100% Modulation on peaks) open-loop switching amplifiers have a reputation for excellent sound quality. I can only comment on Amplifiers based on the TI Solutions. They are excellent, but nothing I'd trade a good tube amp in for...
Moreover, I think it would be very useful if there was a concerted and determined effort to address the weaknesses of current switching Amp's (thanks to energy efficiency legislation from the Europe to Kalifornia that will soon make sure it is illegal to sell any other kind anyway), in the same way a determined person systematically examined and addressed a wide range of problems in Class AB/B Transistor Amplifiers around two to three decades ago...
Ciao T
I could not think of any application where I would want that in speakers. No Speaker I ever designed was able to operate with textbook crossovers unless massive amounts of heavy handed EQ where added.
Textbook crossovers (active or passive) just do not work with real drivers.
And it is by far smarter to modify the filter to do as much of the EQ and response shaping required in combination with the filtering (especially if doing it analogue, be it passive or active, but also in DSP as it saves calculations or DSP horsepower) then to daisychain textbook filters and textbook equalisers 20 stages deep.
Rapidly heading there. Honest.
Well, using TI's all Digital Amplifiers at -24dB attenuation (which is pretty common for normal listening) beats this comprehensively, unless you place the volume control after your crossover, to preserve the SNR, in which case you can better the TI Amp's by a little bit. And I would not claim the TI solution to be state of the art (it is meant for multichannel AV Receivers).
Modern DAC Chips easily measure lower levels of noise than those that you quote, the key to preserving it is good design of analogue stages, if one insists on using analogue amplification.
Now this would be a worthy topic, especially in the light of the ton's of ultrasonic output modern converters output and the tendency of Op-Amp's to have rapidly rising distortion with rising frequency (which you very aptly illustrated in your WW articles), which has quite dramatic implications that normal analogue design works tend not to sufficiently address.
But to be honest, it really seems SO 20th Century to do it that way. If we really want to do old-tech, where are the tubes?
Some specific users with specific performance requirements or prejudices may disagree with the all digital approach (e.g. me, I also disagree with using Op-Amp's of course, or transistors unless absolutely unavoidable), but for mainstream use, even in extreme High End contexts the all digital approach makes tremendous sense (even on performance grounds).
And there are enough "analogue" circuit tricks in switching amplifiers and similar areas that could be gainfully documented (I still remember the little trick of yours to shift the signal take-off for the Miller Compensation Cap to the Amplifier Output which makes a whole lot of difference with not a single component added) in combination with the techniques for "non-textbook" crossovers (and how to synthesise them from the driver response) that would make a book on active speakers very relevant in the 21st century.
Actually, to beat that would be a little non-trivial, due to analogue stage issues that are parallel to those in your crossover (noise, mainly(, but not in any way unsolvable or even really difficult.
Throw together a PCM1794 and a single OPA1632 per output in a competent way and you have equalled if not exceeded the performance you quote at LF and trashed it at MF & HF, at very modest cost.
Well, arbitrarily low distortion does not guarantee good sound and relatively high levels of distortion (such as found in tube amplifiers) do not guarantee bad sound.
Given that arbitrarily low THD is trivial to achieve using modern solid state circuitry (as in "more advanced than the 1950's 'Lin' Amplifier") I find it hard to get excited about low THD either.
The subject of distortion and sound quality warrants more investigation, as in the end it is the percieved sound quality that matters.
I have noticed that some of the faster switching, lower power (meaning they run very close to 100% Modulation on peaks) open-loop switching amplifiers have a reputation for excellent sound quality. I can only comment on Amplifiers based on the TI Solutions. They are excellent, but nothing I'd trade a good tube amp in for...
Moreover, I think it would be very useful if there was a concerted and determined effort to address the weaknesses of current switching Amp's (thanks to energy efficiency legislation from the Europe to Kalifornia that will soon make sure it is illegal to sell any other kind anyway), in the same way a determined person systematically examined and addressed a wide range of problems in Class AB/B Transistor Amplifiers around two to three decades ago...
Ciao T
But you'll have to accept that there are more designers out there than just you. And curiously, they may find something like that useful from time to time. I certainly did for the woofer to midbass crossover in my system. Same with the interfacing of sealed box minimonitors with Butterworth alignments- a 2nd order Butterworth is just the thing for realizing a 4th order L-R to a woofer.
This may not be how YOU want to do things, but it's a big world out there.
(That said, I'm in agreement with you about the rapid obsolescence of analog crossover implementations)
Dear mr. Self,
As much as I have been admiring your work since the late seventies, I have a slight feeling your projected active x/o filter book seems to be a highly academic exercise, so a warning for the average hobbyist is in place I believe.
What is overlooked time and again by many DIYérs is that loudspeaker and filter should be viewed as one. Even on this forum we see all sorts of highly sophisticated electronics with boutique components that are actually completely useless in actual multiway loudspeaker systems, because they only filter the electronic output fed to the power amp and loudspeaker in a textbook fashion, but do not take the loudspeakers acoustic characteristics (SPL measured in box that is) into consideration. The final result is mediocre to poor in most cases.
No matter how wonderful the circuitry or filter topology, what is really needed in all cases is a software program that can optimize the filter curve (and therefore component values) according to a desired acoustic target curve, which sees the loudspeaker and filter as one.
Everyone who has ever tried to optimize active x/o circuitry on the basis of real loudspeaker measurements, will find himself in trouble, unless a lot of experience and knowledge of filter topologies and creativity is at hand. If one wants to look at the complications of properly implementing an active x/o, please have a look a the LEAP tutorials on the website of Linear X. Even with very well behaved loudspeakers, at the end of the day the active filter will become a rather complex affair. The Linkwitz active x/o implementations show more or less the same phenomenon., although it must be added that OB designs complicate matters even more.
Although there is some x/o design freeware available to the DIY er, to my best of knowledge, none of these can handle and optimize active circuitry. Only a handfull professional packages as LEAP, CALSOD-PRO LspCAD-PRO and SPEAKEASY can optimize passive and active circuitry.
It is an interesting and challenging excercise to list and describe all possible active filter topologies, but I am afraid it will not bring the average DIY audio enthousiast any closer to proper active filter implementation, because component values cannot be textbook and/or standardized, but need to be optimized and non standard topologies are needed in most cases.Without having acces to both loudspeaker measurement tools and x/o optimizationsoftware proper active x/o design is near impossible.
In thirty years of publications, both in magazines and later on the internet I have, apart from Siegfried Linkwitz’ designs, seen only one design that was properly executed. This design called “The Force” was published in “Speaker Builder” by one B. Lamy. The acoustic fourth order filters slopes were achieved by using optimized second order standard series feedback opamp sections, buth the real beauty lay in the use of multiple, tunable gyrator treble/bass/cut/lift/suckout circuits for each filter section. Through measurements the gyrators were fine tuned. The whole, very complex, circuit had been optimized in Calsod. This. however, is way beyond what the avarage amateur may dream of, unless he is willing to invest in software and a long learning curve.
Given the flexibilty of DSP and their relatively low cost in recent years(Mini DSP/Behringer/Hypex or PC Soundcards) plus the availability of easy to handle DSP software , it is no wonder they rapidly gain territory over the classic analogue active and passive x/o circuitry in recent years .
Eelco de Bode
.
As much as I have been admiring your work since the late seventies, I have a slight feeling your projected active x/o filter book seems to be a highly academic exercise, so a warning for the average hobbyist is in place I believe.
What is overlooked time and again by many DIYérs is that loudspeaker and filter should be viewed as one. Even on this forum we see all sorts of highly sophisticated electronics with boutique components that are actually completely useless in actual multiway loudspeaker systems, because they only filter the electronic output fed to the power amp and loudspeaker in a textbook fashion, but do not take the loudspeakers acoustic characteristics (SPL measured in box that is) into consideration. The final result is mediocre to poor in most cases.
No matter how wonderful the circuitry or filter topology, what is really needed in all cases is a software program that can optimize the filter curve (and therefore component values) according to a desired acoustic target curve, which sees the loudspeaker and filter as one.
Everyone who has ever tried to optimize active x/o circuitry on the basis of real loudspeaker measurements, will find himself in trouble, unless a lot of experience and knowledge of filter topologies and creativity is at hand. If one wants to look at the complications of properly implementing an active x/o, please have a look a the LEAP tutorials on the website of Linear X. Even with very well behaved loudspeakers, at the end of the day the active filter will become a rather complex affair. The Linkwitz active x/o implementations show more or less the same phenomenon., although it must be added that OB designs complicate matters even more.
Although there is some x/o design freeware available to the DIY er, to my best of knowledge, none of these can handle and optimize active circuitry. Only a handfull professional packages as LEAP, CALSOD-PRO LspCAD-PRO and SPEAKEASY can optimize passive and active circuitry.
It is an interesting and challenging excercise to list and describe all possible active filter topologies, but I am afraid it will not bring the average DIY audio enthousiast any closer to proper active filter implementation, because component values cannot be textbook and/or standardized, but need to be optimized and non standard topologies are needed in most cases.Without having acces to both loudspeaker measurement tools and x/o optimizationsoftware proper active x/o design is near impossible.
In thirty years of publications, both in magazines and later on the internet I have, apart from Siegfried Linkwitz’ designs, seen only one design that was properly executed. This design called “The Force” was published in “Speaker Builder” by one B. Lamy. The acoustic fourth order filters slopes were achieved by using optimized second order standard series feedback opamp sections, buth the real beauty lay in the use of multiple, tunable gyrator treble/bass/cut/lift/suckout circuits for each filter section. Through measurements the gyrators were fine tuned. The whole, very complex, circuit had been optimized in Calsod. This. however, is way beyond what the avarage amateur may dream of, unless he is willing to invest in software and a long learning curve.
Given the flexibilty of DSP and their relatively low cost in recent years(Mini DSP/Behringer/Hypex or PC Soundcards) plus the availability of easy to handle DSP software , it is no wonder they rapidly gain territory over the classic analogue active and passive x/o circuitry in recent years .
Eelco de Bode
.
Dear Pano,
Thorsten touches the point indeed, but the solution he offers (daisy chaining and equalizing) is sub-optimal and far from ideal. Look how LspCad e.a. do it with their optimizers. For most anyone who has ever worked with target curves and filter-optimizers, most other approaches are several steps back.
Eelco
Thorsten touches the point indeed, but the solution he offers (daisy chaining and equalizing) is sub-optimal and far from ideal. Look how LspCad e.a. do it with their optimizers. For most anyone who has ever worked with target curves and filter-optimizers, most other approaches are several steps back.
Eelco
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.