Amplifier stability with very low impedance, high efficiency mid/tweeter section???

The answer is a little more complicated and it lies in understanding exactly what an audio amplifier is.

The so-called power amplifier is actually a voltage multiplier with enough current so that the voltage multiplication remains consistent while driving speakers. In other words, a voltage, not power, amplifier.

And the difference between an ideal amp and a real amplifier is in the output impedance of the amplifier. Think of this as a relatively small resistance that happens right before the amp’s speaker terminal. A lot of otherwise very beefy amps have a lower output impedance at low frequencies rather than high. Usually that’s fine. Most speakers have lower impedance in the bass and this works well.

But take an ESL, which may be 4 Ohms in the low range and 1/3 of an Ohm at the top. The combination of rising output impedance and dropping speaker impedance cause the amp to no longer give a flat frequency response. You lose treble.

Sometimes we hear this problem well in the bass. Take a speaker that drops to 3-ish Ohms in the 100 Hz range. Quite typical for 2 1/2 way speakers or speakers with multiple bass drivers. The less sturdy the amp is, the more the output voltage starts to track the speaker impedance and we are left with a loss in the lower mid bass.

So, best case is that your amp’s output will track the speaker impedance, dropping as impedance drops.  Worst case is that the combination of output and speaker impedance at the top end plus the power requirements cause too much power to be dissipated in the output devices and you'll let the smoke out.

Thanks Erik.  But I'm still not sure I understand and maybe I won't.  Are you saying that driving a high efficiency low impedance mid/tweeter section will be no easier for the amplifier than driving a full range lower efficiency speaker that dips to a similarly low impedance in the bass?  (assume we're talking about cone/dome speakers, not ESLs/panels etc)

Think of this another way.

You have multiple gates you must pass.  You have absolute maximum current and voltage swings which the amplifier can output before failure, or power supply exhaustion.

So you solve those issues by making sure the output never exceeds either.  OK, but then the issue is, is the output flat?  Unless the amplifier has a very low output impedance and the driven impedance is very flat, probably not.

Here is a typical measurement from a tube amplifier:

https://www.stereophile.com/content/doshi-audio-evolution-monoblock-power-amplifier-measurements

Take a look at the black line which is the amp output with a simulated speaker load. As you see, the output tracks the speaker impedance. The ratio between the speaker impedance and the amplifier’s output impedance (i.e. damping factor) determines how pronounced that deviation is.

OK. I understand that speaker impedance fluctuates with frequency.  I think what you're pointing out is that a low dip in a speaker's impedance (no matter at which frequency) is going to present a problem for a typical amplifier to track a flat frequency response because most amps cannot double their current when dropping from 4 to a 2ohm load (and tube amps halve their current).  And that a low impedance can blow up an amplifier even if it's not playing current demanding material (bass), correct?  I also didn't know that an amplifier's output impedance fluctuates with frequency response... 

It's a power and heat sinking problem. The lower the load the amplifier sees the more heat has to be dissipated by the output transistors. Transistors have a maximum operating temperature and they'll fail at that temp so amps are designed with thermal cutoffs well below that limit if the heat sinks get too hot. 

But... if you lower the volume then there is less heat dissipated and if that volume level is loud enough (courtesy of the high SPL of the speakers) and the amp is happy and not overheating at that volume level, then it's not a problem. The amplifier does not care what the SPL rating of the speaker is, that is between the speaker and it's crossover and internal volume -- the amplifier sees an inductive load only.

To give you an idea of the heat sinking requirements:

say I want an amplifier with only one pair of output devices, 50 watts at 8 ohms, biased low A/B, with 36 volt rails. At peak signal input (say 1.2 volts) the power the transistor dissipates at peak is about 45 watts. At 2-ohm loading it's 150 watts. Can't do it, needs more pairs of transistors so the amp is limited in its power output to not lower than 4 ohms.

But if you cut the input voltage with the volume pot down to, say, 200 millivolts -- and that is loud enough for you -- the power dissipated at 2-ohms is 50 watts, which is within the max power spec at 8 ohms.

I understand that the lower the load the more heat has to be dissipated, therefore the more output devices and heatsinking.  I just thought that it was low frequencies that were responsible for the majority of this demand, not higher midrange and up.  

And yeah the amplifier doesn't care about the SPL rating but it does care how hard it has to work.  And a high sensitivity speaker makes it work less hard for the same SPL.  No?

I am old school. If your amplifier cares about the speaker I say look, you're an amplifier, this is your job, tough luck. If you want it easy go back to school, learn how to be a AVR, nobody cares as long as you have a lot of knobs and switches and stuff. But for now, get on with it.

I did lose one to Amplifier Protective Services but that was before I learned to not let them have a cell phone. Take my advice, make an example out of one, word gets around fast, you won't evert have any amplifiers whining about how hard you work them. Trust me. Haven't heard one word of complaint in 20 years at least now.

To understand this better you need to understand what happens when you have an ideal voltage followed by two impedances.  One of the amp, one of the speaker.

V ---> Z(output) --> Z(speaker)  or for simplicity in a DC world:

V--> Ro --> Rs

You need to understand a little about how the varying Z (or R) will affect the voltage at the speaker.

You also need to understand that Zo/Ro is almost never flat, so the amplifier's output impedance is also changing.  The inability of an amplifier to double it's output as impedance is halved is related to this.

So a 100W amp that can't double all the way down to 2 Ohms will also probably not be able to double at 1 Watt output.

Lol MillerCarbon.  

Erik, I've looked at a bunch of other stereophile measurements of amps.  They all fluctuate in impedance but but it's not by very much, at least most of them.  I guess you would need an amplifier with an output impedance of less the .2 ohms across the spectrum and a beefy power supply/enough output deviced to drive a 2 ohm load safely.  

...low impedance of 2 ohms.  Would this be a problem for most amps? 

It can be a problem. My rule is that if the loudspeaker gets into the 2 ohm range I would want to match that with a very robust amp that boasts about its low impedance performance.  

Erik, I've looked at a bunch of other stereophile measurements of amps.  They all fluctuate in impedance but but it's not by very much, at least most of them. 

Yeah, but that's wiht a simulated 4 Ohm speaker. :) Lower speaker impedance = more fluctuation.  I'm not saying it will be unbearable. Just explaining how a more robust amp with a lower output impedance would behave.

In some cases the interaction could be euphonic.

But take an ESL, which may be 4 Ohms in the low range and 1/3 of an Ohm at the top. The combination of rising output impedance and dropping speaker impedance cause the amp to no longer give a flat frequency response. You lose treble.

@erik_squires You might want to leave ESLs out of this since they are not 'voltage driven' loudspeakers, unlike most box speakers. Their efficiency is not mapped by their impedance curve. So you need the same amount of power at 10KHz as you do at 100Hz to obtain the same sound pressure- and these two frequencies will be very different impedances! If you put an amp that acts as a voltage source on a speaker like that it will tend to be bright and lack bass impact as a result. There are exceptions- for example the Sound Lab ESLs have controls and switches on the back to allow you to address this problem.

@poseidon1500 There is not an amplifier made that will be harmed by the low impedance of your midrange/tweeter array. But there is a different issue which you might want to take into account!

All amplifiers make greater distortion into lower impedances. In addition, the ear uses higher ordered harmonics to sense sound pressure and so is keenly sensitive to their presence either natural or generated by amplifier distortion. It interprets them as 'bright and harsh'.

To complicate matters the ear has its greatest sensitivity at higher frequencies: between about 4-7KHz. This sensitivity is often called the 'Fletcher-Munson curve'. Something to keep in mind is that if you have an amp that can double power as impedance is halved (IOW, acts as a voltage source), harmonics of the bass notes will be audible coming from the high frequency array. This will contribute to brightness. Of course this will always be the case, the problem is that with many amplifiers as frequency is increased, there is a certain point where the feedback falls off inside the audio band. This causes the amp to make more distortion and this is very likely at or near the frequencies your MTM array is operating. If you are planning a conventional solid state amp this really sounds to me like a recipe for brightness and harshness.

If it were me I would insist on auditioning the speaker and amplifier combo in your home!

Erik,

Ok that makes sense.

Ralph,

That is what I needed to know, and then some! Thank you!! Could you explain the part about feedback falling off in the audio band as frequency increases? I have no idea what that means.  Something to do with negative feedback diminishing and thus THD going up?

My answer you found humorous was funny yes but intended to make a point and get you thinking. I mean actually really thinking not bandying words about which is all that is going on now. No progress being made towards answering your question, which except for one detail was pretty well thought out.

The "for example" you are talking about, a MTM array used 200Hz and up, is of course what Tekton does. If you want to make this a guessing game going round and round nobody ever really getting anywhere because nobody really knows what you’re talking about, you don’t need me you got the right guys on the job already.

This is the point where usually I would give some deal clinching examples of precisely why they are the right guys for that particular job. But not today. If you want to know why it is no problem at all for your amp or any other, in fact is an advantage inherent in the design that helps explain why they sound so doggone good, let me know I will fill you in.

MC, 

I didn't just laugh at your post.  I definitely get what you're saying.  To a large degree I agree with you.  If it won't blow up and it sounds great then that's all I care about.  I know the principle behind Tekton and why people like them.  If you want to tell me more though I'd be happy to hear what you have to say.  

Have you thought about a new class D amplifier? Very good on low impedances. Almost any amplifiers with MOSFET outputs should work fine.

Could you explain the part about feedback falling off in the audio band as frequency increases? I have no idea what that means.  Something to do with negative feedback diminishing and thus THD going up?

You stated it correctly here. When the distortion increases above a certain frequency, you can count on harshness as a result. Combine that with an amp that easily doubles power as the load impedance is halved, driving a speaker where the high frequency array might be only 2 Ohms... that sounds like its going to be bright and harsh to me.

The technical explanation of why the feedback reduces at high frequency is quite technical! Its related to the gain of the amp circuit (with no feedback) and also its overall bandwidth (again with no feedback). That is why its called Gain Bandwidth Product. At low frequencies almost any amp has the GBP to support its feedback and  that is one reason solid state amps tend to have good bass. But at some higher frequency the GBP falls off, preventing the feedback from being the same as it is in the bass So the distortion increases. Amp manufacturers are a bit reluctant to publish this information since it tells you a bit about how the amp might sound.

All I'm saying here is that you have to be careful about the amp you select to work with a speaker like this! Put another way if the speaker were to have the usual 8 Ohms in the MTM array it would likely sound smoother for no other reason than the amp driving it will have less distortion.

Ralph, thanks for explaining that!   This is simply for a speaker I'm thinking of making myself, not for an existing product.  Once I model it I'll have a better idea of the impedance curve.  But that certainly gives me something to think about.  Thanks again! 

fiesta75,

I quite like my YBA and haven’t heard a class D that matches it in overall tone so I’m sticking with it but I may be using ICEpower for 200hz and down.

Just have a good amp with substantial power supply and stable to 4 amp double power into lower ohms preferred 

my coda has a Massive 3kva potted transformer doubles down to 2 ohms 

and 120 amps short term take 2-16 amp slow blow fuses .

my brothers Gryphon Diablo doubles power to even under 2 ohms 

the coda is a steal at $5500 or less ,and 3 power choices  when you order 

upto 18 first watts in class A

 This is simply for a speaker I'm thinking of making myself, not for an existing product.  Once I model it I'll have a better idea of the impedance curve. 

@poseidon1500 To prevent the array from being low impedance you can put the midrange drivers in series rather than parallel. The efficiency does not change although the sensitivity does. A lot depends on how you design the crossover and the selection of the midrange drivers but that becomes a simple engineering issue. It would allow the speaker to sound smoother (all other things somehow being equal...) regardless of the amp used.

Audioman58, 

I've never listened to a Coda amp but would love to!

Ralph, 

I should say the main reason I want to use two midrange drivers is to match their output level to the tweeter for the same given power input.  But I'm going by the notion that two drivers in parallel will result in a 6db boost in output.  That is 3db from the doubling of drivers and 3db from the doubling of power.  From my understanding two drivers in series will result in the same output as one because voltage for each will be cut in half (-3db) and the doubling of drivers will bring it back up again by 3db.  I also don't know the exact explanation between efficiency and sensitivity other than one is expressed in percentage and the other in SPL/power/distance. 

Either way, last night I modeled the drivers in a typical parallel crossover and it turns out the lowest they dip is 3 ohms.  Not as bad as I thought.  

Ralph (atmasphere)- Thank you for all you clear and concise info. 

Question: What about the potential use of autoformers to improve impedance matching? ,,. e.g., Paul Speltz' autoformers:, quoted here (not a recommendation, I I have no connection or financial interest, just an example):: 

Of course, more wires & connectors, etc.

Thoughts? 

Thanks, 

Ken

There is another, very serious issue that no-one has mentioned.  An amplifier will often - usually - run into stability issues when it is loaded by a reactive load.  An ESL looks like a capacitor to an amplifier, and that means it’s stability margin is affected.  This can cause oscillation at worst (frying things), or at the least sound degradation.  At low frequencies, speaker drive units are non-linear, which means that the impedance they present varies according to what they happen to be doing at the time.  The effect is that they can act as a large inductor, which again can drive an amp into instability.  When an amp does go unstable, it’s output will become momentarily independent of the input, and it can take a little while for it to get itself together again afterwards.

Generally it is better to run an amp into an ESL with some series resistance or inductance such as 6 feet of zip cord to avoid these issues.  These issues may also be the cause of speaker cables sounding differently - the reactive aspects of the wiring are not normally considered.

Back in the 70’s it was these issues addressed by Julian Vereker  that made the Naim amplifiers so exceptional.  A 60W amplifier capable of providing 8A was unheard of (and is still very unusual).  Nowadays amplifiers such as the Sanders Sound Magtech are designed to avoid reactive load problems as of course they have to work with his ESL’s.