Completely impressed and saddened by DIYaudio

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Have this overwhelming desire to say to so many members how impressed I am with the energy and expense you have gone to in the pursuit of audio. Cheers and accolade to those many of you. It is truly amazing!!

I am equally saddened with a long experience in audio the same errors are being made with the same faulty premises found at the time of my college days in the early 70's. I am so surprised in these 40 years since then little change has occurred, little clarification and almost no corrections of models to the actual physical facts with mathematical simulation now the substitute for laboratory efforts. So now the expert is the guy with a computer instead of a laboratory.

I commend almost all of you and feel sad for you at the same time. Keep on trying though. Am sure some will come upon amazing results which well satisfy their needs like Apexaudio has for his PA work and shared it freely with all. A good thing.

I shall ponder what is learned here for a long long time. Cheers-:cheers:
 
I am equally saddened with a long experience in audio the same errors are being made with the same faulty premises found at the time of my college days in the early 70's. I am so surprised in these 40 years since then little change has occurred, little clarification and almost no corrections of models to the actual physical facts with mathematical simulation now the substitute for laboratory efforts.

Examples?

CMIIAW, but I figure that ALL lab research is predicated on theoretical models - mathematical or otherwise.

In my own puny way, i follow that line - I set out some design parameters, sketch out a block diagram, work up suitable schematics, model them using the known parameters, adjust for best overall performance, build, test, adjust to fine tune, listen.

Generally , the better my modeling, the less twiddling later.

Since the lab efforts of much more able practitioners are often reduced to mathematical models, I can bypass a lot of the early lab stuff. Surely thats a good thing?

With so many people doing so much in so many different directions in this field alone, let alone associated fields, it seems to me that the reason you have a perception that little has changed is perhaps because there is little to change - that there is nothing new under the sun, and the models that you perceive as somehow lacking actually represent the physical reality with a fidelity that is good enough for even the more critical and educated engineers and listeners.

Otherwise, someone would have done it.
 
Just another Moderator
Joined 2003
Paid Member
Hi sumaudioguy, how could it be any different? I would postulate that the large majority of people posting or reading at diyAudio have no formal training in Electrical engineering. Most would be here because it is a hobby. Many of them will have excellent knowledge gained through research, and experimentation, many of them will be just starting out in this wonderful hobby and have little or no knowledge or experience. There are of course very knowledgeable people with formal training here as well. This can cause friction, especially if their is intolerance of the views of the less informed (a bucket that I will firmly place myself in).

With that in mind, is it not to be expected that many people will make mistakes (and hopefully learn from those mistakes).

As an example take myself:

Over the last few days I have been trying to design a low pass filter using an FDNR. I've been using LTSpice to help me. Everything looked ok (or at least I thought it did) response out to 20K was almost identical to a passive realisation but then deviated (I wasn't too concerned about that as the corner freq was 200Hz). When doing a transient analysis it showed up that the circuit was oscillating at about 80Khz (this is something I would not have been able to detect if I actually had built the circuit as I do not have a scope).

On trying to work out what the problem was (doing a lot of reading) I think that it was the fact that there was a phase reversal after 20Khz that was probably causing the oscillation. I've changed the circuit, the phase reversal is gone, and so has the oscillation. I don't know (or understand) why what I have done has removed the phase reversal, but it worked and I wouldn't have been able to detect (or fix) the problem without the use of a simulator. I now have a much better chance of building a working circuit. Without the simulator, I may never have built the circuit in the first place, as I did not have enough of an understanding of how the circuit worked to be able to progress as far as I have. Iether that or I would have built it, it would have sounded terrible, and I would have given up as I wouldn't have the test equipment to find the problem. Yes I know that even if it simulates ok, it may not work in real life, but the chances are much better.

I will probably post my circuit sometime in the next few weeks (perhaps before, perhaps after I actually have built it). Maybe someone will find it interesting, or possibly even useful. Maybe it will be completely flawed due to my lack of formal training, but I will most certainly not be making any representations as to its fitness for purpose. I may give my own subjective impressions, and probably some objective measurements.

The question is, should I post such a design when I have done so from a starting position of very little theoretical knowledge, no formal training, and just the aid of a simulator to get to where I am? I think that provided I make that clear that there should not be any problem with that.

If I published my design and said it was the best active crossover ever and criticised everyone who begged to differ then I guess that would be a reason to be saddened, but hopefully what I have related above gives an insight into my thinking, and my reasoning as to why a site like diyAudio will always have the potential to have people making the same old mistakes based on the same faulty premises :)

Tony.
 
When ever I was in college doing research in loudspeaker design in the department of physics the professor and I had decided pretty much everything there was to be know about audio was known by 1940 and forgotten by 1960. This is just a thought though.

I understand the modeling pretty well as an early user and pioneer in perfection of models so I agree the simulations are very helpful but do not at all automatically lead to understanding. Understanding the effects and putting that into a big picture is what I had hoped long ago would happen with time. This clearly has not.

The example of wintermute is great in that he got his circuit working. On the other hand without throughly testing the circuit in the lab that circuit could have unexpected behaviors missed by the simulation.

An example is there is a string right now simulation the Adcom555 amp. It is an oscillator. I have proved it time and time again with many different units with old time storage scopes and "snap shots" of the waveform as the amp breaks into oscillation and then stops while driving a typical cone loudspeaker. That fact is not showing in the simulation.

Another example is the notion a typical speaker is a load. Sorry, wrong. A voice coil loudspeaker is an active source of signal. Any sound is detected by the cone and causes the cone to move thereby driving the amplifier. To model a voice coil loudspeaker simply as a load is naive and always done.

This list could go on and on and on. These are the realizations understanding brings and will never be part of a model until someone puts them into the model.

Another is the common misconception a voice coil loudspeaker has but one resonance. For the moment let me just talk about woofers with resonance in the 10-60Hz range. In reality there is an electrical resonance defined as the current and voltage being exactly in phase with one another as measured at the voice coil. Then there is the mechanical resonance which is defined as the exact peak in the impedance curve as measured at the voice coil. I have never seen these two be the same in testing over 1000 drivers. Sometime these are very far apart but typically the mechanical resonance will be 35Hz and the electrical resonance will be 38Hz. I have seen these different in tweeters by as much as an octave- no kidding! There is no model which includes such behavior. This leads to gross errors in box design of course. Something almost every box designer has experienced who has measured the results as this effect causes gross errors in the Small method.

I can do this all night but believe I have made my point.

When I have talked about these things or written papers and submitted them for publication all that occurs is rejection. It just does not go with the popular idea of the time. Remember please, the earth used to flat and the universe circled the earth and it was heresy to say otherwise punishable by death.
 
Last edited:
Ex-Moderator R.I.P.
Joined 2005
Sometime these are very far apart but typically the mechanical resonance will be 35Hz and the electrical resonance will be 38Hz.

Well, if thats all it is then Im less worried
I always take for granted that the real Fs is higher than specs
Sure, it could mean phase issues
But I dont see how I can change that

Im more confused about why its to be expected that specs are off

And how often do we say "if only..."
Why is that we know what we want, and the designer often doesnt
Seems like we actually know quite a lot more about speaker design, than the many of those who design the drivers

Nothings perfect, but what can we do about it
We are not in control of the whole world
We can only hope to make the best of the options we have

You want us to understand something
Its not that I dont want to, but I still dont understand what it is you want me/us to do about it:confused:

You salute us, thats great
Maybe theres even more going on than you have experienced since january, or there have been
Not every day brings new things to the world
 
When I have talked about these things or written papers and submitted them for publication all that occurs is rejection. It just does not go with the popular idea of the time. Remember please, the earth used to flat and the universe circled the earth and it was heresy to say otherwise punishable by death.

I don't think anyone is going to put you to the sword here.

I DON'T have an EE background, but I'd characterise the issues you have highlighted as being at the finesse end of the spectrum. That is, they do not produce gross errors or even errors that are detectable by the general population. Rather, they are theoretical considerations that MAY have application in very extreme circumstances.

In this your points are not dissimilar to protests over the commonly held descriptions of atomic structure where atoms are described as mini solar systems, electrons regularly orbiting a solid nucleus of protons and neutrons. We have known for quite some time that this is not actually the case, but the simplified concept not only appeals but works at the level that it is being applied.

By and large, that is the realm that we operate in here. In accounting terms, the minute details that you describe fall outside the values that are material to us - the errors they produce are negligible by comparison to the errors produced by other issues (including ourselves!), and in any event, they are not easily controlled.

Cheers!
 
Well, if thats all it is then Im less worried
I always take for granted that the real Fs is higher than specs
Sure, it could mean phase issues
But I dont see how I can change that

Im more confused about why its to be expected that specs are off

In this example the result will throw the popular Theil/Small calculation for Q off by 100% because of the error from the theoretical Theil impedance curve. That wrecks box design. Real Q is say .34 and measure from the impedance curve it is .77. that is no small error and not at the edge or finesse. This also makes the simulations worthless because the data entered are so far off.

I have gone as far a providing manufacturer data and other proofs of what I have learned and truly most of "learned engineers" have blown a fuse. Pretty much all I get is "you do not know what you are talking about." Even with data supporting my claims produced by third parties. Yes some are very interested but the "experts" are dogmatic in both their arrogance and ignorance. This does not help the true DIYer either.

I see this simply as these experts believe that all things audio have been figured out and it is there job to make certain their (very limited) knowledge is correct and held high. To even the casual observer this is not true so... this can only be an ego thing, or lack of self esteem. Some of the non-experts have ask really good questions and even some have made very good observations. To many reject the new knowledge or information which leads to threads of nothing but disagreement where the novice can gain nothing.

Really, I am not into minute detail. Gross errors like oscillating amplifiers and measurement that are off by 100% really muck up the audio design process and worse, the resulting sound defines a new level of awful. Any amplifier that has oscillation while listening is not even an amplifier but an oscillator with amplification. How does that help a DIYer? And it really confuses so many things at every level leading to 40 years of chaos.

On the other hand someone as myself can glean some new perspective especially from some of the real discoverers and some of those fantastic efforts by the guy who has no idea so he is open to anything. To that end I have enjoyed my times here and suppose some others have also.
 
frugal-phile™
Joined 2001
Paid Member
In this example the result will throw the popular Theil/Small calculation for Q off by 100% because of the error from the theoretical Theil impedance curve. That wrecks box design. Real Q is say .34 and measure from the impedance curve it is .77. that is no small error and not at the edge or finesse. This also makes the simulations worthless because the data entered are so far off.

I for one would like to learn more about that and how to separate the 2 resonances out...

dave
 
How to measure a woofer

No Problem but you need a scope, DVM with good frequency response, a good oscillator, and at least a 50 watt power amp. I use a 250 ohm 10 watt series resistor and some sort of adjustable power supply that can be a large battery and a 10 watt finely adjustable resistor. Take all readings to as many significant figures as are available. I use a frequency counter for frequency set for period measurement in microseconds.

After the driver has been "broken in" make sure it is in some sort of normal mounting with the frame vertical and solidly held like drivers are normally used except mounted in free air. To get the mechanical resonant frequency hook up as I have shown. In this sim the 4 ohm resistor represents the driver. Turn up the oscillator until like 20 volts is coming from the power amp and adjust the frequency until the scope signals on both sides of the resistor are exactly in phase. This means current and voltage are in phase and here you have the electrical system resonant frequency. Next watch only the voltage across the driver ( channel B in this sim) and very carefully peak the voltage (maximum voltage) on channel B by adjusting the oscillator. Here you have found the mechanical system resonant frequency.

If you want to really know the driver the moving mass needs to be measured. Here a bigger amplifier may be needed. I like to measure mass at 1/2 xmax using the same set up. This get the measurement into a "real life" amplitude and not the silly 1mW figure so many use. Who listens at a 1mW anyway? Headphones?

Adjust the oscillator to achieve an amplitude about 1/2 Xmax. Here I use a white dot on a dark cone so I may see the amplitude and measure that with a quality aluminum scale. The amplitude is needed later. So now the amplitude is known and the frequency of mechanical resonance should be used on the oscillator and the voltage across the driver should be at maximum. Turn the drive off and add mass to the cone. On woofers I use 100 grams typically of that easy to remove but stick clay like stuff for posters or modeling clay. Now turn the setup back on and make a rough tuning for the maximum voltage across the driver. It is very important now to carefully measure the amplitude of motion and adjust the oscillator drive level down such that the same amplitude is achieved as with no mass. The amplitude must be the same if the measurement is to be accurate. Now carefully find the maximum voltage across the driver again and recheck to make certain the amplitude of driver motion is the same. This is the new mechanical resonant frequency of the driver. So you now have Fs electrical, Fs mechanical, and Fs with mass added(F2). From the two mechanical measurements and the known mass added (m2) the moving mass and compliance may be calculated.

____________ (2*pi*F2)*(2*pi*F2)*m2
Moving mass = --------------------------------------
_____________ (2*pi*Fs)*(2*pi*Fs)-(2*pi*F2)*(2*pi*F2)

1/C= 1/(2*pi*Fs)*(2*pi*Fs)*m1

So now you have moving mass and compliance and mechanical Fs.

The only thing left is the BL product. Unfortunately there is no good way to determine BL other than by direct measurement. I use a surface plate and place the driver magnet down and put a dial micrometer probe onto the point on the driver where the dust cap meets the cone and set the dial to zero. I use a dial ticked of in .0001" increments for accurate measurement. With the dial reading known set that same 100 gram mass of clay on the dust cap which will cause the driver to move down and the dial caliper to read differently. Now pass quality very smooth DC current through the voice coil until the exact same reading is seen on the dial caliper before the mass or current was added. The DVM should be in the circuit so current is accurately known. I usually take this reading over and over until I see very little error as sticktion or vibration can affect the setup. In between each reading I re-zero the setup while disconnecting and current source so no zero error is made. With a little practice and patience a good set of readings here can be made. So now you have a current and a mass. F=ma so F=9.8meters per second (gravity) time the mass of 0.1kg which is .98 newton meters. Current is what the meter says. Force factor is F=BLi so BL= F/i This is BL.

One more thing is needed which is Re. It is very important the Re of the driver is taken in such a way no motion of the cone is likely. In my lab I turn every moving thing off including HVAC, have everyone leave, and turn the driver with the cone down against the bench and very carefully measure Re with a 4 wire ohm meter. Now you have every thing about the driver that is important.


____2 pi Fs m
Q= ----------------
____Rm + Φ²/Re


_____ Φ²S²
Ra = ---------- * (.0005610)
______Re m²

This is the point where I forget that I needed Rm.

Rm=(Bl*Bl)/(Zs-Re)

Zs= the impedance that is measured up there when determining the mechanical resonant frequency. To get Zs use the peak for mechanical resonance and measure the voltage across the resistor. Current is then V/R=I which gives I (current) through the resonant system tuned to mechanical resonance. Now measure the voltage across the driver while tuned to mechanical resonance. The voltage Vres/I=Z at resonance.

Now you have all the parameters for the mechanical system. The electrical system is only really important in making crossovers. It is the mechanical system that get tuned in a box. You may now use whatever model you like to tune the driver in the box. Make sure any fudge factors are turned off in the model such as the silly box loss and so on as those are all about bad measurement and really have nothing to do with real design.

I have used this method for 35 years and have yet to have a design be more than 3% different from theoretical when measured response is tested. Using the Small method of course errors can exceed 100%.

A lot of this should sound fairly familiar. You will find driver specs from manufacturers of often off quite a bit because everyone uses the impedance curve and the points not at the peak to estimate a whole lot of things. Because the electrical impedance curve is not a good analog of the mechanical system the errors can become quite large. Please notice these are are direct measurements focused on the mechanical system and not inferred from the electrical response.

Ra is the asymptotic response which is the driver efficiency in acoustic watts. 1% is 94dB and not 100dB because the constant is set for 1/2 space.

A few other things that are really handy is in a sealed box, raising Q by 2 say from .35 to .70 exactly doubles the resonant frequency as found in the box. Also for any kind of port, vent, or passive radiator ideal coupling is achieved when when the free air Q is = .383. If a crossover is used or a long speaker wire which both raise Q then a lower free air Q is needed. This new Q may be calculated above by adding the expected added crossover resistance to the Re resistance and finding the new free air Q. Qs much different than this do not work well in vented boxes even though there are many papers describing how to tune them. Further, it is extremely important to realize any signal from the vent in a correctly tuned box is one full cycle behind or delayed from the signal at the active driver. This means if the vent is tuned for 30Hz the vent signal is 33.3ms delayed and stops 33ms after the drive signal on the active driver. This is darned close to echo and really gives that double bass ba-bump sound where only bump should be. Tuning to 25Hz is 40ms of delay is should be considered echo. So this means ported, vented, or passive radiator systems should never be tuned in the region of 20-40Hz because of this delay problem. 40Hz and above is 25ms or less and tuning 20Hz and below puts the signal out of the audible range and is therefore no problem.

That is it and it works every time no matter the driver. SURPRISE
 
DIYAudio to me says it all. Have fun! There are other forums to be really serious about EE.
When they come up with a piece of equipment (AP System 17+?) to measure the THD+N on a violin I promisse I will buy it because I do not have the "Golden Ears"
Cheers, Elk

Funny you should mention the violin. That is the instrument studied when I was in college. Several of those papers were published but my work on that did not get me co-authorship.

Yes have lots of fun!
 
Administrator
Joined 2004
Paid Member
<snip>

That is it and it works every time no matter the driver. SURPRISE

This is very cool stuff, not only is the premise well explained, but the technique to do it is clearly explained and I am confident will demonstrate the issue clearly. I have long suspected that the electrical resonance and the mechanical resonance (even as reflected electrically) in drivers are two separate entities, and wondered at the more or less total lack of discussion about this issue.

I am one of those people who has been known to use parallel resistors across 16 ohm compression drivers in order to flatten out the impedance variation with frequency. As an EE that made absolute sense to me - I had way more efficiency than required for the application so trading some in return for a flatter driver impedance at the XO was a good trade off. It worked extremely well for me. (Also my preferred amps are not particularly good voltage sources.)

When it comes to speaker design I am pretty much of a newbie. I've designed one, and built a few other designs. I have a lot to learn, and enjoy reading your posts - they're thought provoking.



Kevin
 
Last edited:
Reasons to be cheerful

You're Impressed and Saddened, huh?

Well I'm Proud, I'm Happy and I'm Thrilled.

Get over it. There are much worse things out there than a few inaccuracies in models. How 'bout an entire section of the community incapable of distinguishing between science and fantasy and another whole section only too happy to pander to them as long as it's of personal financial benefit? ...to say nothing of the uncontrolled indulgence of obsessive-compulsive disorder. If this were an anorexia sisterhood, there'd be calls to shut us down...

Other things, however, have improved over the period you cite.

The semiconductor replaced the valve.

Solid state memory supplanted 78's.

w

A white LED? What's an LED?
 
Loudspeaker modeling, Thiele/Small etc

I believe there is a problem with the Thiele/Small model & parameters commonly used to represent loudspeaker drivers.

My suspicion was aroused when I noticed that published impedance curves do not agree with the published voice coil inductance (Le) at frequencies well above resonance, where one would expect the VC impedance to be almost purely inductive.

It's not just that the values don't agree - the slope of the impedance curve is not what one would expect from an inductor of ANY value.

A bit of Googling revealed the reason: VC inductance is not a "pure" inductance, but rather a "lossy" inductance, due to frequency-dependent eddy current losses in the magnetic circuit.

Thus the "Re" and "Le" values given in the TS parameters for drivers are, at best, a crude approximation of reality, and only valid over a limited frequency range.

To save a thousand words (and because I'm not an expert), here's a couple of papers that describe the effect and it's causes and propose better models :
http://users.ece.gatech.edu/mleach/papers/vcinduc.pdf
http://www.tymphany.com/files/papers/AES122nd-Impedance.pdf

A couple of quotes from the first paper:
When the series resistance is separated and treated
as a separate element, it is shown that losses in an inductor require
the ratio of the flux to mmf in the core to be frequency dependent.
For small-signal operation, this dependence leads to a
circuit model composed of a lossless inductor and a resistor in
parallel, both of which are frequency dependent.
...it was noticed that
the phase of the impedance of drivers approached a constant at
high frequencies after the series resistance of the coil and the
motional impedance term are subtracted. For a lossless voicecoil
inductance, this phase should be 90 degrees. However, experimentally
observed values were usually in the 60 to 70 degree range.
and a quote from the second paper:
Comparison between the impedance curve predicted by the simple equivalent circuit and the actual measured impedance reveals some differences. First, at fmin the impedance magnitude should be very close to RDC, and the phase angle should be close to zero. In practice, however, the impedance at fmin is always higher than predicted. Typically for a woofer, the impedance is often more than 1 dB higher and the phase is not zero. Secondly, the measured slope of the impedance curve at high frequencies is typically closer to 3 dB/octave rather than the 6 dB/octave we would expect if the only inductive element in the circuit were a conventional inductor. Obviously, a box simulation based on this simple equivalent circuit model will often result in significant errors.
This ties in nicely with sumaudioguy's measurements and observations and seems to give a reasonable explanation for why impedance minima and maxima do not correspond with zero phase.

Regards - Godfrey
 
Last edited:
From Godfrey- You have one explanation and here more-

The air trapped between the spider and the top plate is another mechanical resonator. The air trapped between the dust cap and the pole piece is yet another. Both of these get reflected into the electrical circuit. And yet there is more as frequency increases the coupling between the voice coil and the diaphragm or cone changes which changes the reflected impedance up there. Then adjacent winding starts showing its interwinding capacitance (especially true on 4 layer VC) and so another UFO is tossed into the mix. All this is why I abandoned the reflected electrical analysis because it just got SO complex.

The dust cap one on a 2 inch voice coil is like about 400Hz. So vented pole pieces but then the vent has a tuning which changes it again...ouch!


As a note I have never seen the electrical resonance be at a lower frequency than the mechanical one.
 
I did some driver measurements at university using a gain/phase analyser and it was apparant that voltage and current were not in phase at resonance, or in fact over quite a large area of interest. Not much thinking later and some basic principles and this was immediately obvious given that you are driving an inductor.

Now the task becomes to obtain proper current/voltage measurements without spending thousands on test kit.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.