I share your opinion on this.
And I understand the circumstances well. I've written a few songs myself, and when it comes time to record a demo version for friends and family, I have the usual conundrum: I don't own, or can't play, some of the instruments I want on the track.
Drums are the most obvious example. I owned a set of electronic drums at one time, and took a few months of drumming lessons. But that was when we still lived in a house. Today, I live in an apartment, surrounded by quite a few fragile and very elderly people. If I want drums on a home recording today, they're going to come from either a Digitech Sdrum, a Digitech Trio+, or drum samples programmed into a sequencer or software drum machine on a Linux computer.
Years ago, I wrote a 'cello part for one of my songs. I never recorded that song, in part because of the 'cello problem; it's not an instrument I can afford to buy, store, or learn how to play. If I ever make a recording of that song, I'm going to have to either play the 'cello sounds with a keyboard, or program them into a sequencer. Or leave out the 'cello entirely, which would be a huge shame.
I don't think it's a bad thing that we have the option to use software generated instruments, but I do think it has helped push real, physical instruments closer to extinction, and I think that is sad.
-Gnobuddy
Concerning peak pulse distortion. I have to admit that my investigations were in the early nineties, and memory maybe as reliable as orignal floppy disks from that era. I think that peak pulses were quite short - but your explanation of the physics points into the opposite direction. On the other hand - I have never been a very fast player but more one who likes to "chew on the notes".
I understand. True "Hi-Fi" was achieved decades ago, certainly by the time the CD player came along, eliminating the last audible deficiencies in the entire audio chain (except for microphones and loudspeakers). Then most of the actual scientists and engineers left the field,, leaving it wide open for the subjectivist wackos to take over. The lunatics have been running the asylum for decades. Now we read utter nonsense like this:
"Wire is directional!"
"Putting my amplifier on spikes improved the pacing of the music!"
"Digital audio files produce staircase waveforms that sound bad!"
It's quite true that 10% THD consisting of mostly 7th harmonic, will sound very different than 10% THD consisting of mostly 2nd harmonic.
But if THD is sufficiently low, then no human ear can hear any of the distortion, no matter which harmonic it might be. In practical terms, if THD is below 0.1% from 30 Hz - 15 kHz, then no human ear can hear it. Doesn't matter if it's 2nd harmonic, or 7th. You still can't hear 0.1% THD.
(I have never read of a properly conducted double-blind listening test that found people could detect anything below 1% THD. So the 0.1% THD specification is already a full order of magnitude below audibility.)
These days every cheap little chip amp and class D board can achieve that, and usually beat it by an order of magnitude. Even cheap little amps are audibly perfect for Hi-Fi, as long as you keep them out of clipping...
But that is too simple a reality for some people to believe, so endless arguments will continue, until all of us old geezers are pushing up daisies, and everyone still alive listens to music using tiny earbuds connected to the always-on digital spyware device they carry around with them.
This is an interesting argument, and it has been around for many, many decades. It sounds plausible, but is it correct?
There is an incredibly simple experiment that proves beyond doubt if the concern is valid, or not.
The experiment is this: tap off a bit of the amplifiers output, divide it down to the same amplitude as the input signal, subtract one from the other using a differential amplifier, and look at the resulting signal. Do this while the amplifier is hooked up to real loudspeakers (full duplex voltage generators), and playing actual music.
Let's call the amplifier output signal (adjusted to same amplitude as input) A. Let's call the input signal B.
If the amplifier is perfect, the output is an exact copy of the input; A=B, therefore (A-B) equals zero.
If the amplifier is misbehaving in some way, the output is not a true copy of the input. Instead, it is a copy of the input with some sort of added distortion. Let's say the output A equals the input (B) plus some distortion (D).
This time, (A-B) equals (B + D - B). In other words, the difference signal is just the distortion, D.
Peter J. Walker of QUAD fame did this experiment decades ago - in the 60s or 70s. Even back then, he found that, with good amplifiers of the time, the output from the (A-B) experiment was buried in the noise floor of the amplifier. Whatever distortion there was, was so small that it couldn't even be extracted from the faint background hiss of the amplifier.
Walker did find some amplifiers where this did not hold true, most particularly, amplifiers that put a huge electrolytic capacitor in series with the loudspeaker output. (This was common in early solid-state Hi-Fi power amps, which ran off a single power rail, rather than the symmetrical +/- supply rails that came along later, when Hi-Fi took lessons from op-amp integrated circuit design.)
But when's the last time you saw a solid-state power amp that used a 1000uF electrolytic cap in series with the speaker output? I think it was in the early 1970s for me (I was still a kid).
Decades later, our solid-state Hi-Fi amplifiers have gotten even better. Today the entire power amp is direct-coupled, except for maybe one input cap, so there are no electrolytic caps to charge and discharge and distort the signal. There are negligible phase shifts at the bass end of the spectrum to confuse the subtraction process, too. Also today's transistors are much faster, so there's less phase lag at 15 kHz to confuse the subtraction. Nowadays, even cheap Hi-Fi amps pass Walker's test. Input minus output equals zero (limited by the noise floor). Ergo, output equals input. The amplifier is doing exactly what it should do, in spite of being connected to those weird full-duplex voltage generators.
Note that the difference signal (A-B) can be above the noise floor, and still be far below the human ear's ability to hear it. If the difference signal is, say, 60 dB below the music signal, it is already far, far, far below the threshold of detectability. (Auditory masking: https://en.wikipedia.org/wiki/Auditory_masking )
These days we have audibly perfect audio signal sources and amplifiers: no more wow, flutter, rumble, print-through, tape hiss, record surface noise, ticks, pops and thumps. We have ruler-flat frequency response over the entire audio frequency range. THD far below audibility. Watts/dollar has fallen so far that we can have more power than we could ever need for extraordinarily low prices.
There are only two items in the entire audio chain that are still audibly flawed: the microphone, and the loudspeaker (and the room the speaker interacts with.)
This is engineering reality. Audio electronics is perfect, done, dusted, nothing more to see, nothing more to do. Loudspeakers are still very audibly imperfect - no two models ever sound exactly alike, meaning there are audible differences, meaning there are audible imperfections.
But if I wandered over to the Hi-Fi forums and expressed these opinions there, I would probably be tarred, feathered, and ridden out on a rail. Maybe stoned to death, I dunno. It would be ugly.
I used to be interested in Hi-Fi electronics when I was young. I'm not interested any longer. There's nothing new left to learn there.
E-guitar electronics, on the other hand, is still lots of fun to tinker with. And there are constant opportunities to build myself something that doesn't already exist as a commercial product.
For example: I have been using a Mustang Micro as my amplifier for our online music jams since the pandemic shut down in-person jams. The tiny Fender Mustang Micro headphone amp packs some very good sounds in it. But you are stuck with one sound throughout your song; the tiny buttons are too small to see, and too awkward to find with your fingers if you can't see them.
So I designed and built a Mustang Micro channel switcher. Simple enough in concept: a table-top box with input and output jacks for two Mustang Micros, and a relay that switches between the two outputs.
Make the box run off the usual 9V power supply, so you can power it just like all your other guitar pedals.
Now build in USB power to keep the Mustang Micro's charged.
Finally, add a footswitch that controls the relay. Now you can set one Mustang Micro for clean tones for the verse, set the other for overdrive for the solos, and switch between them on the fly with your foot in mid-song. Happy happy joy joy! 🙂
Right about the time I finished buying my second Mustang Micro, and designing and building my Mustang Micro channel switcher, I discovered the Flamma Preamp. It's cheaper than one Mustang Micro, never mind two. And its already foot-switchable between clean and dirty channels. It's a much less awkward form factor, too. And it already runs on standard 9V FX-pedal power. And it's less fragile. And the controls are big enough to see. Doh.
The only remaining advantage to the Mustang Micro switcher is that the Mustang Micro has onboard effects (the Flamma Preamp doesn't). So I can set up one Mustang Micro with, say, clean tone and spring reverb for rhythm playing, and set up the other one with a nice overdriven tone, along with delay and large-room reverb, for singing single-note solos or crunchy chorus sounds. One toe on the foot-switch is all it takes to go between them.
-Gnobuddy
I wonder if it is the same goat?
https://southpark.cc.com/video-clips/1a7z0b/south-park-looks-like-we-re-going-to-afghanistan
https://southpark.cc.com/video-clips/1a7z0b/south-park-looks-like-we-re-going-to-afghanistan
Well I've done simulations where replacing an ideal 8 ohm load with a simple LCR circuit for the speaker was enough to sabotage the amplifier's 'internal' THD by >10x or 20dB due to heavy phase shifts. If we assume that a phase-shifted load is a 'flaw', then yes, we could exonerate just about any amplifier, but if we accept that 99% of speakers are the way they are, then at least some responsibility falls on amplifiers to cooperate with what they've got.
The THD issue can probably be dealt with in much the same way that box vibrations (dB scale) can be reduced by adding mass to the box (linear scale). I.e.: brute force, just throw more NFB at the problem. Flattening the speaker's impedance using unobtainium electrostatics could also help. But even a reasonably sized polypropylene cap for passive zobel filters can get pricey.
Moreover, I learned from experimental results that a speaker could usefully serve as an active bass trap. By adding 10 ohm in series to a 10" woofer in a 40L slightly 'over-damped' box, a prominent room resonance @ 40Hz was significantly reduced, and the parametric EQ I was using for room correction had to have its notch dialled back by a few dB and Q reduced (2.5-ish --> 1.5-ish, IIRC). A very positive result when I was expecting the exact opposite.
In hindsight, it makes perfect sense to me that a "holy grail" of perfectly controlled cone position would have the fatal side-effect of causing the cone to be a 'node': a region of minimum velocity and maximum pressure. So it acts like a one-way valve, adding energy to the room according to its own measurable perfection, but then utterly fails to cushion the output. If I had to design an active bass trap from scratch, I'd probably start with a big box + transducer etc. So why not use the ones already there?
There's a 1989 paper (among others) claiming 10+ dB improvements in speaker distortion by regulating the current instead of voltage. But even then, I guess they were battling decades of momentum, and nearly every speaker manufacturer I've seen tunes the tonal balance to favour the readily available voltage amplifiers. (With a couple of PA and instrument brands as possible exceptions e.g.: Eminence, with the bass resonance conveniently positioned below roll-off, slyly favouring tube amps).
As further confirmation, nearly every tweeter routinely gets "padded" down a few dB, and people often don't seem to realise that the difference in sound is at least partially a reduction in distortion, not just EQ. If EQ makes a difference in subjective harshness then there must be something else going on.
OTOH, another paper warns against too-high output resistance because of jump distortion. The frequency response may have more than one result, depending on prior conditions. The results may be sudden glitches where there are pre-existing variations in amplitude between adjacent frequencies.
So now I'm at a stage of looking at a tube-like system with a moderately high output resistance, preferably adjustable.
I'm looking at 2 alternative topologies that could work. A modified Howland current pump adds a sense resistor in front of the load. And it would also favour NFB even in tricky cases like headphones with a common ground. But it kind of seems like yet another Rube Goldberg machine when open drain (open collector) gives a transconductance amplifier directly.
The THD issue can probably be dealt with in much the same way that box vibrations (dB scale) can be reduced by adding mass to the box (linear scale). I.e.: brute force, just throw more NFB at the problem. Flattening the speaker's impedance using unobtainium electrostatics could also help. But even a reasonably sized polypropylene cap for passive zobel filters can get pricey.
Moreover, I learned from experimental results that a speaker could usefully serve as an active bass trap. By adding 10 ohm in series to a 10" woofer in a 40L slightly 'over-damped' box, a prominent room resonance @ 40Hz was significantly reduced, and the parametric EQ I was using for room correction had to have its notch dialled back by a few dB and Q reduced (2.5-ish --> 1.5-ish, IIRC). A very positive result when I was expecting the exact opposite.
In hindsight, it makes perfect sense to me that a "holy grail" of perfectly controlled cone position would have the fatal side-effect of causing the cone to be a 'node': a region of minimum velocity and maximum pressure. So it acts like a one-way valve, adding energy to the room according to its own measurable perfection, but then utterly fails to cushion the output. If I had to design an active bass trap from scratch, I'd probably start with a big box + transducer etc. So why not use the ones already there?
There's a 1989 paper (among others) claiming 10+ dB improvements in speaker distortion by regulating the current instead of voltage. But even then, I guess they were battling decades of momentum, and nearly every speaker manufacturer I've seen tunes the tonal balance to favour the readily available voltage amplifiers. (With a couple of PA and instrument brands as possible exceptions e.g.: Eminence, with the bass resonance conveniently positioned below roll-off, slyly favouring tube amps).
As further confirmation, nearly every tweeter routinely gets "padded" down a few dB, and people often don't seem to realise that the difference in sound is at least partially a reduction in distortion, not just EQ. If EQ makes a difference in subjective harshness then there must be something else going on.
OTOH, another paper warns against too-high output resistance because of jump distortion. The frequency response may have more than one result, depending on prior conditions. The results may be sudden glitches where there are pre-existing variations in amplitude between adjacent frequencies.
So now I'm at a stage of looking at a tube-like system with a moderately high output resistance, preferably adjustable.
I'm looking at 2 alternative topologies that could work. A modified Howland current pump adds a sense resistor in front of the load. And it would also favour NFB even in tricky cases like headphones with a common ground. But it kind of seems like yet another Rube Goldberg machine when open drain (open collector) gives a transconductance amplifier directly.
In other words, you want any of Papa Leonidas Fender's early tube guitar amplifiers. They all have exactly the flaws you are looking for. 😀
I learned long ago not to trust my own unsupported ideas. The human mind is fantastic and conjuring up ideas that usually have no relationship to reality. Trusting in those ideas without experimental verification and solid mathematical support, leads to centuries of arguing about how many angels can dance upon the head of a pin. ( https://en.wikipedia.org/wiki/How_many_angels_can_dance_on_the_head_of_a_pin? )
In this case, it seems plausible to think that a loudspeaker is somehow screwing up the output of a modern Hi-Fi amplifier. It is a perfectly reasonable hypothesis; but a hypothesis must be tested before we believe it, otherwise we are in danger of believing something that isn't true.
In this case, the hypothesis was tested decades ago, by Peter Walker, and many others since. Rod Elliott even had a write-up and test circuit on his website for doing this kind of "subtract amp output from amp input and look at the difference" testing.
And when the hypothesis is tested, it turns out that amplifiers with high NFB and low output impedance actually do a wonderful job of controlling their output signal, while playing actual music (not simple sine wave test signals), while connected to actual loudspeakers with all their supposed misbehaviours.
The experiment showed that the amplifier isn't misbehaving. The accusation was false. The amplifier has been exonerated. Reality simply does not support the seemingly-plausible hypothesis that the amplifier output "goes funny" due to speaker back-EMF.
When carefully controlled experiments fail to support a hypothesis, the only honest thing to do, is to discard the hypothesis. That hypothesis was wrong, plain and simple. To make progress, the wrong hypothesis must be discarded.
Science is full of examples of this. For instance, in the 1800s, physicists believed in a mysterious medium called the "aether" that filled all space, and through which light travelled.
In the late 1800s, Albert Michelson and Edward Morley designed a very clever, very sensitive experiment to verify the predictions of the aether hypothesis (which predicted that the speed of light would vary when measured first in the direction of the earth's motion around the sun, and then at right angles to it.)
When Michelson and Morley ran their experiment, however, their results flatly contradicted what they had been expecting. The speed of light was exactly the same in every direction. Aether did not behave the way it was supposed to. Most likely it did not exist.
Scientists had to throw out their cherished and long-held belief in aether. A young man named Albert Einstein heard about the results of the Michelson-Morey experiment, and postulated that the speed of light was, in fact, a universal constant. That led him to develop one of the most brilliant bits of science in the history of physics - his theory of Special Relativity.
More on the Michelson-Morley experiment here: https://en.wikipedia.org/wiki/Michelson–Morley_experiment
Subsequent history has verified a thousand times over that Michelson & Morley were right, and so was Einstein.
I worked in a loudspeaker design group at one time, for a company that made pro-audio equipment. Our speaker design experts padded down tweeters only when the tweeter had a higher sensitivity than the midrange / woofer crossing over to it. This is a very common situation - speaker manufacturers deliberately design their tweeters in this way, precisely because it is easy to pad down a tweeter without losing much power as heat, while padding down a woofer or midrange would be very inefficient.
In other words, padding down the tweeter is done to produce a flat frequency response, not to generate EQ or somehow eliminate harshness. If not padded down, there would be a "step" in the frequency response at the crossover frequency, with the tweeter's range of frequencies elevated above the midrange.
High output impedance from the power amplifier is also a bad idea, which is why audio engineers began trying hard to lower it, many decades ago, even in the era of tube Hi-Fi. One of the things that made audio engineers happy with tube "ultralinear" stages was that the output impedance of a pentode is very high, but when you apply the local negative feedback that makes it an ultralinear stage, the output impedance drops tremendously.
High amplifier output impedance is a bad idea because it reduces electromagnetic damping at the woofer's fundamental resonance frequency, resulting in a bass peak, and lumpy, woofy, poorly controlled bass.
More precisely, the Q of the driver fundamental mechanical (bass) resonance rises from a low of Qts towards a typically much higher value of Qms, as the amplifier output impedance is increased. Qms is invariably far too high, producing a huge hump in the frequency response, more like a wah-wah pedal than a Hi-Fi speaker.
High amplifier output impedance also makes the amplifier frequency response vary with loudspeaker impedance. For instance, the rising impedance of the speaker voice coil generates unintentional rising treble response in Leo Fender's (tube) guitar amps. It's part of the sound of Leo's guitar amps, but this sort of far-from-flat frequency response is a terrible idea for Hi-Fi audio reproduction.
In Leonidas' guitar amps, the lumpy bass response was made much less audible because of the poor bass response of the open-back speaker cab, really just a folded baffle.
All this is getting further away from "instruments and amps", and I'm not sure we will agree on any of it. So I'll stop here, though I'm happy to continue discussing issues specific to solid-state guitar preamp design.
-Gnobuddy
For a solid state guitar amplifier that sounds tube-like, I think we should be looking in the direction of simplistic, relatively low gain designs like the Nelson Pass Zen amp or the french creations by Jean Hiraga and PLantefeve. Getting rid of (most of) the feedback they are using will lead in the right direction of solving the main problem, how to make them clip tubely.
It is interesting that the Zen amp includes the one thing that Peter Walker found is guaranteed to make an amplifier fail the "output minus input equals zero" test - a huge, whopping, electrolytic capacitor in series with the loudspeaker!
But for a guitar amp, yes, I agree. No feedback, low transconductance, and smoothly curved transfer characteristics are all good things. That is not enough to make it sound like a "tube", but at least it will keep away harshness, and that is a good start.
Interestingly, a Russian engineer who goes by "KMG" on the Internet, pretty much nailed the problem of making MOSFETs behave like vacuum triodes. See his posts in this thread: https://www.ssguitar.com/index.php?topic=2039.0
In a nutshell, KMG lowered MOSFET transconductance by having up to five semiconductor diodes inserted between MOSFET source and ground. He also arranged another diode to gently clamp positive-going input signal swings, which has the result of softly clamping negative-going signal swings at the MOSFET drain. This emulates the soft saturation characteristics of the half-12AX7 triode.
KMG also used a high-voltage MOSFET, and ran his solid-state circuit at the same B+ voltage as the original tube stage. Basically, he matched nearly everything - waveforms, overload margin, everything - to make his MOSFET stage behave very nearly like the real vacuum tube circuit.
I believe KMG did make one mistake, though. He used a 100k drain resistor on his MOSFET stage, to match the 100k anode resistor used for the (half a 12AX7) vacuum triode. But the vacuum triode has an internal anode resistance of only 70k, so the output impedance of the tube stage (with 100k load) is actually about 40k.
MOSFETs, on the other hand, have very little slope to the Id-Vds curves, like a pentode. Output resistance of the LND150 MOSFET is much, much higher than 70k. As a result, the output impedance of KMGs MOSFET stage is close to 100k, rather than to 40k.
This is not an issue when the output is very lightly loaded. But what I found on the test-bench is that if you try to drive a heavier load, such as the typical Fender TMB passive tone control circuit, then KMG's MOSFET stage doesn't behave the same as the vacuum triode amplifier he was trying to emulate.
Simply using KMG's methodology, with a 39k drain resistor instead of a 100k one, and adjusting B+ voltage and the number of Schottky diodes connected in series with the MOSFET source, might do the trick.
The only real downside to KMG's awesome work, is that his tube emulation circuit is relatively complex and uses a number of components to work its magic. I think this has kept many hobbyists from tinkering with it.
KMG had sound clips of his designs posted on his website, and to my ears, they sounded very, very good. Certainly the best analogue solid-state guitar amp sounds I can recall hearing.
At the other end of the complexity spectrum, I remember stumbling across a circuit (and sound clip) of a very simple circuit, a small-signal JFET in common-source mode, driven by a low gain, non-inverting op-amp stage. You could say it was nothing more or less than a small-signal version of Nelson Pass' Zen amp, without a speaker. 🙂
At least to my ears, that very simple circuit produced relatively sweet distortion for notes higher up the guitar neck. Low-frequency notes were buzzy and uninspiring. (But the circuit didn't have bass roll-off, which is invariably used in "proper" guitar distortion circuits. I think proper bass roll-off might have resulted in better sounds.)
-Gnobuddy
If you only want to control voltage to the nearest uV and show that it's correct then sure, a test designed to confirm that NFB does what it does, will do that. But the point remains: if the speaker has known performance issues like non-linear inductance, which, as a lumped parameter appears in series (modulating the actual voltage that's seen by the sound-producing part) then using a voltage control scheme is still incorrect no matter how linear it is for perfectly resistive loads.
Unless we want to start designing new speakers from the ground up to suit a pre-existing amplifier? I can actually think of a couple of things that could be explored, like swapping out the conductive iron cores for non-conductive "iron powder". The saturation would still be imperfect and prone to 'flexing' in response to the temporary magnetic fields set up by the voice coil, but at least the problem of non-linear damping caused by eddy currents would all but disappear.
I'm not worried about lumpy bass. Notch filters or Linkwitz's circuits can correct for it within reason.
Until they clip. Which, with speakers with misbehaviours, will happen pretty much anytime they get within 10db of rated output (as the internal signals are going to be a crap load bigger). Even worse if there's resonance or time delay in the speaker misbehaviour. Worse if the output device actually has a high impedence which is being "fixed" by feedback.
THD is a blunt instrument at best (which we've known since before Crowhurst, N. H., 1957 Some Defects in Amplifier Performance Not Covered by Standard Specifications JAES 5(4)) and we know that the right "crap" is audible 60db down (Bareham, John R. "Automatic Quality Testing of Loudspeaker Electroacoustic Performance." SAE Transactions (1986). On the other hand most speakers can't manage 1% THD at reasonable listening levels but apparently they're perfect.
At the end of the day you need to match the speaker (and its flaws - see Wolfgang Klippel or Ragnar Lian or Lynn Olsen or Walt Jung, to taste) to the amplifier (and its flaws) to the room, the db level and the program material. And your own ears & personal triggers.
(There's a beer here abouts called "Old man yells at cloud". Some days I resemble that remark 🙂 )
That's why the usual approach to SS guitar amp design falls short I think. The idea is to do all of the sound shaping in the preamp, and then simply feed that into a plain ordinary hifi-ish power amp, with lots of gain and feedback, wich you prevent from clipping itself. But a great part of E guitar sound is exactly the clipping of the output stage and its interaction with the speaker! Somehow it has become accepted that it is not worth it to look into output stages other than the ordinary class AB or, nowadays, class D stuff.
Yes, I think we agree entirely that speakers are most definitely imperfect devices!
IMO, amplifiers are already audibly perfect. Speakers are definitely not audibly perfect.
To me, it doesn't make sense to take a perfect amplifier, and mess it up in the hope of making a flawed device (speaker) perform less badly.
So yes, I'm of the opinion that it's the speakers we need to make better. Make the flawed thing better, don't take the perfect thing and make it flawed.
My long-ago speaker designer colleagues knew a lot about making speaker drivers better, but ultimately, trying to make a loudspeaker cone do exactly as instructed by the voice coil voltage, is like herding cats.
So one of the research projects I initiated while working at that job, was to re-visit a wonderful old idea that brilliant Phillips engineers cooked up five decades ago: motional feedback. Sense the woofer cone motion, apply negative feedback to the amplifier, so that the cone motion is the thing that is corrected, rather than the voltage (or current) waveform to the voice coil.
This was in the late 1990s, long before the era of MEMS silicon accelerometers. I built my own one-axis accelerometer using a piezo ceramic disc in a housing I machined on the lathe in the basement, used it to replace the dust-cap in a spare woofer, and started a little research program to figure out what electronics (loop-shaping) I would need to produce stable negative (motional) feedback.
In the end, I was able to get 18 dB of negative feedback. I saw 3rd-harmonic distortion from the woofer (driven hard at low frequencies) fall nearly 18 dB in response. The close-miked frequency response was ruler-flat from 10 Hz to somewhere north of 500 Hz, pretty much regardless of what size sealed box you mounted the woofer in. Bass was "tight" in a way I'd never heard before.
And that turned out to be a problem. In listening tests, some felt the bass was too tight.
The problem was unfixable: it turns out recording and mixing engineers put blankets in the bass drum and EQ the drums and bass until they sound right when played though monitors that do not have motional feedback; those monitors unavoidably have some bass "boom" and hangover built in, so the engineers in the recording studio actually end up creating a signal where the bass is too tight, but their boomy monitors restore it to the right level.
When fed into my MFB woofer, however, the "too tight bass" on the recording was mercilessly revealed.
Twenty five years later, you can buy cheap and accurate MEMS accelerometers with integrated preamps, so the job of making DIY woofers with motional feedback would probably be a lot easier.
-Gnobuddy
Texas Instruments was selling a 600-watt (!!!) class D amplifier board for $75 USD a while ago.
We have now reached the point where you can buy so much amplifier output power for so little money, that for domestic use, only the truly insane will turn up their amp until they hear it clip. Anyone who tries it will experience PSH (permanent shift of hearing, i.e. permanent hearing loss) long before clipping occurs.
Maybe that would be a good thing, as they won't be able to hear the police knocking on their front door after the neighbours file noise complaints. 😀
-Gnobuddy
A musical stealth amp is beginning to formulate in my head....
I think a missing piece of the puzzle here is impedance conversion. It sounds like you were basically making a compression driver with high motor strength, except that if the horn it was connected to was shaped more like a living room and less like an anechoic outdoor venue, it would've produced ringing peaks and nulls.
People's hearing seems to vary a lot when it comes to interpreting bass resonances. So it may have sounded perfectly dry, and any unexpected rattles would be the room, and not the speaker 'upstream'.
A sanity check could be attempting to invoke the same room resonances with an acoustic double bass.
A plucked string may produce large, low-amplitude forces strong enough to bend wood, and then the wood pushes the air. But the leverage is duplex, so the vibrations in a room are less likely to become standing waves.
There have been a slew of hybrid guitar amps sold by Vox and other companies over the years, most containing one lonely 12AX7 surrounded by a sea of BJTs. None have gained a great reputation for sounding really good.
(Not surprising to me, I don't think one or two lonely triodes are sufficient to get the same "tubey" characteristics as an entire guitar amp.)
Electric guitar players with lots of money have been doing as you say, with their Kemper Profiling Amps and Axe-FX boxes, for years. No valves anywhere, just DSP chips and lots of MOSFETs.
Some years ago there was a British company who stuck a valve (tube) preamp and a powerful solid-state class D power amplifier in one box. I forget the brand name. I don't know how well the product sold.
For living-room e-guitar playing and practice, I have been doing much the same thing, in a flea-watt way, using a Flamma Preamp and a little Lepai class D desktop power amp driving an old pair of Hi-Fi "mini component" speakers rescued from a thrift store.
I took our weekly music jams online early in 2020 after the pandemic shut down all our lives. For those, I use a Fender Mustang Micro, with its (stereo) outputs run into a little Yamaha mixer. My wife and I both sing, so our vocal mics also go into the same mixer. The remaining stereo input on the mixer is fed by either a drum machine or a looper pedal, which I use from time to time when I get sick of the limitations of having only one guitar and two voices.
The little Yamaha's mixer's main output goes to what is essentially a flat-response powered speaker, so we can monitor our vocals and my guitar.
Meantime, the mixer's headphone output is suitably attenuated, fed into a generic "USB guitar cable", and that in turn plugs into the little Raspberry Pi computer we use for the online jams. Linux sees the cable as a USB sound card, so you can feed the incoming audio signal into Zoom or Skype or whatever video chat service you prefer.
The Mustang was bought before I ever heard of the Flamma Preamp. Otherwise, I would probably have gone with the Flamma instead.
I have heard this often, but struggled to find supporting evidence from my ears. Perhaps because I never play at ear-bleeding loudness levels.
I have heard guitar-amp output transformers reducing bass response, sure. I've heard them reduce bass response dynamically, with increasing output power, too. But I've never heard any other hint that the transformer was affecting the sound in any audible way when clipped.
So I dunno about the output transformer, but output valves certainly affect the sound dramatically, whether overdriven, or not. Some of them cold-bias themselves and generate alarming amounts of crossover distortion when driven hard, and in a rock-music context, the raucous sound of crossover distortion has been welcomed by some.
For me, the contribution of the output valves to "clean tone" is much more important. Beam tetrodes (like 6V6 and 6L6) produce lots and lots of 2nd harmonic distortion. This is quite clearly audible, particularly in single-ended output stages. The attached excerpt from an old General Electric 6V6 datasheet shows 12% THD from a single 6V6 (pushed to 5.5 watts output from a 315V B+).
The operating conditions are not far from Leo Fender's Champ, and there is no doubt that 12% THD is audible to anyone without cloth ears.
By comparison, from studying the datasheets and looking at load lines, the ubiquitous 12AX7 small-signal triode pair tends to produce considerably lower distortion. From the guitar amp point of view, this points to the fact that the output valve plays a bigger part in the guitar amp's tone than the preamp valves do.
Using a push-pull output stage mostly cancels out all even-order harmonic distortion, which is why the same datasheet shows more output power, but lower THD, for two 6V6 in a push-pull output stage.
SE or PP, since the valve power amp definitely does contribute to the guitar sound: if you wanted to, there's no reason why you couldn't build both a valve preamp, and a flea-watt tube power amp, with the output of the whole thing fed to a class D solid-state power amp.
I designed and prototyped a roughly quarter-watt all-valve guitar amp a while ago, so this approach is entirely practical. You don't have to waste tens of watts of heat in the valve amp, or use big, hot, bulky output valves.
The tricky part is finding suitable output valves for a flea-power amp. The usual suspects (EL84, 6V6, 6L6, etc) are orders of magnitude too powerful for this job. They also need lots of power, take up lots of room, generate lots of heat, and cost lots of money to replace.
My first attempt at low-powered valve guitar amp design used a pair of 6AK6 output pentodes, from the days of American "tube radios". In push-pull, output was around 2 watts RMS. An order of magnitude less than most "small" valve guitar amps, but, as it turned out, still an order of magnitude too much power to use in an apartment. It was perfect for our live jams, though. (We don't have a drummer.)
For the later 250 mW amp, I used a little triode-pentode valve for the driver and output power, single-ended, running at a few mA of current and only 135 volts B+. Lots of careful tinkering with the valve datasheet led to an operating point that would work with a 25k load impedance, which was obtainable from an off-the-shelf Hammond output transformer.
Push-pull operation with the same tiny pentodes, would need a 50k transformer primary, not available off-the-shelf, but used by Vox in one of their amp designs (they used the two triodes in a 12AX7 as a push-pull pair of "output" valves).
In fact I originally intended to have my 250 mW amp operate in either stand-alone, or in Randy Bachman / Gar Gillies "Herzog" mode, feeding a bigger guitar amp, or a solid-state clean amplifier. I never actually used it in the latter mode, though.
For anyone unfamilar with the Herzog: In a nutshell, guitarist Randy Bachman put his e-guitar through a Fender Champ, complete with 6V6 and output transformer, then fed the output of the Champ into the input of a bigger Fender guitar amp.
The story is that Bachman loved the sounds he got, but both amps failed very quickly. He took the failed amps to "Gar" Gillies to have them fixed. Gillies found a way to stop the Champ from blowing up (basically, he just added a resistive dummy load to replace the Champ's missing speaker).
Gillies website says he and Bachman later developed the concept further: http://garnetamps.com/zog_story.htm
I think a hybrid amp is an entirely legitimate concept, but it's not what this thread started out to be...
-Gnobuddy
NOOOOO! I detest the sound of compression drivers and the horns they are connected to.
Motional feedback is nothing like a compression driver. You put an accelerometer on the voice coil itself, which generates a feedback signal proportional to the actual motion of the voice coil. That signal is fed back to the amplifier input.
The amplifier will now generate whatever output signal is necessary to make the voice coil do exactly what it is supposed to do.
You're not relying on constant voltage drive (which works pretty well), or on constant current drive (which works very badly).
Instead, you have a fast-acting servo feedback, which constantly senses and corrects the voice coil motion. Not the magnetic flux, not the generated force, but the actual movement.
The result is that the actual voice coil motion exactly tracks the audio signal fed into the amplifier input, subject to the inevitable limitations imposed by physics and available drive power.
At subwoofer and woofer frequencies, the entire speaker cone operates as a rigid piston. Applying motional feedback to the voice coil means you get (nearly) perfect cone motion. And nearly perfect sound-waves are generated at the speaker as a result. (Launch them into a bad room, and yes, the room will mangle them.)
Without motional feedback, on the other hand, voice coil motion follows Newton's law: f=ma. Unfortunately, the force "f" comes from many things, including nonlinear speaker surrounds, nonlinear voice-coil spiders, and back-emf from the voice coil that doesn't exactly track cone motion if the voice coil begins to move into regions of lower magnetic field strength on large cone excursions.
Properly designed, MFB works wonders on woofers. Remember, I measured nearly 18 dB reduction in 3rd harmonic distortion from the woofer at low frequencies. That is an astonishing improvement.
Most of that distortion was caused by the loudspeaker's surround and spider, both of which have "s-shaped" curves of spring force vs voice coil displacement. That curve generates mostly 3rd-harmonic distortion.
But the "s-curve" is the only way to keep the speaker from self-destructing when driven too hard. The non-linear spring forces attempt to gracefully brake the voice-coil to a halt before it flings itself entirely out of the magnetic gap.
No doubt. So I had as many co-workers as possible come and audition the MFB woofer. Some liked it. Some preferred "soggier" bass.
But it became evident that, on many recordings, the sound on the recording was "too tight". Most likely because the sound was not mixed with MFB woofers in mind. It has to sound good on a boombox, you know...
There were no rattles. I made sure of that!
The density of air is about 1.2kg per cubic metre at STP (standard temperature and pressure). The density of spruce (used in the top of a double bass) is in the range of 400 - 600 kg per m^2.
The wood is three orders of magnitude denser. As a result, the wood is much more effective at driving the air, than the other way around.
This is why upright basses make very poor microphones.
Not really duplex. More like one-way, but with a very tiny leakage the other way.
I find the sea of loosely-attached cheap plastic sheet on the face of a Fender Stratocaster, acts a bit like a thicker version of the plastic sheet that makes up a banjo head. This is why a 'Strat tends to be relatively loud when played unplugged (no amp).
The slab of dense maple that makes up the top of a Les Paul is much less willing to vibrate. Played unplugged, a Les Paul is much quieter than a 'Strat.
-Gnobuddy
But you will have clipping going on already at much lower levels. You know Eguitar has a a huge dynamic range, transients can be 20 times as high as the average signal level. So if you are playing at 1 watt average level, when you hit the strings hard a 15 watt amplifier's output stage will clip.
I don't know much about class D yet, but read somewhere that it is impossible to have a high output impedance. Since guitar speakers are normally designed for high output impedance tube amps.. I do think one should regard the amplifier and speaker as one system here, because the sound depends on how they interact.
I guess the question is, how much of that transient remains after the e-guitar signal gets through the (tube) preamp. The more you overdrive the preamp, the less of those transients remain.
There are four different types of negative feedback. Two of them are based on how the feedback signal is fed back to the input (in series, or in parallel, with the input signal).
The other two depend on whether you sense the output current, or the output voltage.
If you sense the output voltage and feed that signal back, it has the effect of lowering the output impedance.
If you sense the output current and feed that back, it has the effect of raising output impedance.
Like you, I don't know enough about class D power amps to be sure if the last option is possible or not. But at first sight, I can't think of a good reason why not. Usually the trick is to put a very small resistance (say, 0.1 ohm) in series with the loudspeaker, and sense the voltage across that, which is proportional to speaker current.
Agree, if you use a guitar-specific speaker, then you need to feed it from the (high) impedance it expects.
Many solid-state guitar amps do this, and still sound like garbage. Clearly, high Zout doesn't fix the other deficiencies of typical SS guitar amps.
The downside of this approach being that you have to change out a very expensive component (the guitar speaker) to create any substantial change in tone.
The alternative approach is to use a flat-response speaker, and put the "cab simulation" (really EQ that mimics the guitar speaker frequency response) into the amplifier.
This gives you enormous flexibility with the guitar tone. But of course, like all engineering choices, it has its own downside, because flat-response speakers are invariably far lower sensitivity than guitar speakers, so you need a ton of power if you want insane e-guitar loudness to compete with a loud drummer on stage, et cetera.
Fortunately, the cost of a tone of amplifier output power keeps falling, and falling, and falling. A couple of years ago Texas Instruments was selling a 600-watt class D amplifier board for $75 USD. You could get a suitable switch-mode power supply to feed it for $40. Throw it in a box and a switch and a volume knob and a fuse and IEC inlet, and you could probably still come out under $200 for everything. That's an insane amount of power for the price.
There are top level pro guitarists who just use an an Axe-FX or Kemper run into the (flat frequency response) P.A. system for live gigs.
The little $76 Flamma Preamp is the poor man's substitute for the Axe-FX or Kemper. Speaker/cab simulation is built in, and the output is designed to be fed to a flat-response powered speaker.
I can't think of any reason why you couldn't feed that output into a high-powered P.A. system if you were playing a live gig with a large audience. For me, stuck in my living room in a world shut down by COVID-19, all I need is any small flat-response powered speaker.
Flamma also sells just a cab simulator by itself, which might be an excellent match-up with little DIY solid-state guitar preamps, or tube ones, for that matter - you don't need to add a $150 guitar speaker, and then mic that, to get your preamp to sound right: https://www.amazon.com/FLAMMA-FS07-...&keywords=flamma+cab+simulator&qid=1644088512
-Gnobuddy
That's not all that different from using a very powerful magnet and dropping the Q to very low levels. You just used a different technique. The difference is that makers/users of compression drivers have taken the next logical step: noticing that the 'torquey' motor can push a column of air longitudinally, and transfer energy into the air more efficiently.
The trouble with horns arises when there are reflections and part of the energy returns to the cone, while the cone is simultaneously applying force in the opposite direction (and/or it has been made somehow immovable...). This is a recipe for a dramatic increase in resonance. Mechanical delays have basically turned negative voltage feedback into positive feedback, with an alternating mix of positive or negative depending on frequency.
Notice that with current drive, the resonances will be different because the rear chamber and cone mass, as well as damping, are now part of the system. Now notice that by varying the output resistance, two competing sets of resonances could be made to cancel out.
For a woofer, a typical living room is just a badly designed bass horn. They are parts of one system. The box usually starts off well with a 180 degree conical section, but then it's a mess of 90 degree turns and furniture.
As for the usual complaints about high output impedance, like: voice coils going into thermal runaway, or the bass resonance and tonal balance being different than what people are used to, none of that is insurmountable.