SET amp owners thread

[DaveC]

When I bought my Audio Tekne system Imai told me to “be careful” because he has eliminated much of the high freq distortion. That it won’t sound loud to me and I may have the tendency to keep turning it up until I adjust to it.

IMO, this is a key quality of a good system. You should not be able to tell how loud it's playing as a result of distortion levels.

This is also an issue with drivers... while many folks like very light paper cones distortion increases with excursion at too high of a rate vs other designs. My midrange driver uses a very light paper cone and it took a lot of effort to solve this problem! An example of this is the "shout" many of these drivers exhibit when excursion increases.

So this phenomenon is not only an issue with the amp, it could also be an issue with the speaker.

 

[Atmasphere]

If the listener is sitting not much more than 3 meters from the speakers not much power is needed to hit high 90sdb; which is loud enough for most music and listeners. I would argue < 5 watts for peaks and < 1 watt for most listening.

A friend of mine had McIntosh MC-501 monos for a long time and he had first quite demanding KEF Ref 207s and later quite demanding Thiel CS3.7s. He also liked to listen loud (louder on average than all but one of my hifi buddies) and I never saw the needle go above 50 watts on the KEFs and above about 20 watts on the Thiels. Neither was nearly as sensitive as the Daedalus speakers. We were sitting probably 3.5-4 meters at his place.

As to distortions making it sound loud, it is only the high order harmonics that cause this perception. As long as the amp is not clipping the harmonic products for a SET should rise in concert together and that effect will not occur. In amps where low orders are suppressed the rise of high orders will be more noticeable and add “edge” to the sound. SPL also plays a major role. We are less sensitive to higher order harmonics as the SPL increases, so as long as the pattern of distortion is maintained the sense of purity will be retained. That pattern needs to be both even and odd harmonics because that is a pattern from nature and our ear/brain has mechanisms to mask as the ear generates its own harmonics when receiving and transmitting sound waves. Jean Hiraga noted this preferable pattern (he called it monotonic but he meant a smooth exponential decay in modern scientific terminology) quite a long time ago. An interesting experiment by Keith Howard where he added distortion to digital files found that an all odd harmonic distortion pattern was the most offensive sounding, indicating that little to no masking of high orders is occurring with such a pattern. The best, unsurprisingly, was no added distortion but second best was all orders with an exponential decay with increasing order. All even (not really possible I think without a computer program) orders was worse but better than all odd.

That meter, unless it was set up to read peaks, probably was not telling the whole story! The peaks may well have been 2-4 times more power! You might need 5 watts for peaks; the problem is that as the higher orders increase in amplitude, even though masked by the 2nd and 3rd, the ear still uses them to gauge sound pressure. Keith's experiment didn't include the masking issue so probably isn't relevant here.

You can simply parallel a bunch of drivers and increase efficiency as much as you want, but this doesn't mean a 1.5W 45 SET is the best answer if you have a dozen 15" woofers involved.

Just to be clear by adding more drivers in parallel you increase sensitivity (since the load impedance is reduced while the drive voltage is unchanged) but efficiency is unaffected (since 1 watt is evenly distributed between all the drivers in the array). Assuming of course that sensitivity is 2.83volt/1 meter and efficiency is 1 watt/1 meter....

With tube amplifiers the efficiency spec is more useful since tube amps don't generally double power as the load impedance is halved.

 

[morricab]

That meter, unless it was set up to read peaks, probably was not telling the whole story! The peaks may well have been 2-4 times more power! You might need 5 watts for peaks; the problem is that as the higher orders increase in amplitude, even though masked by the 2nd and 3rd, the ear still uses them to gauge sound pressure. Keith's experiment didn't include the masking issue so probably isn't relevant here.



Just to be clear by adding more drivers in parallel you increase sensitivity (since the load impedance is reduced while the drive voltage is unchanged) but efficiency is unaffected (since 1 watt is evenly distributed between all the drivers in the array). Assuming of course that sensitivity is 2.83volt/1 meter and efficiency is 1 watt/1 meter....

With tube amplifiers the efficiency spec is more useful since tube amps don't generally double power as the load impedance is halved.

McIntosh meters are apparently pretty accurate…they claim 95% to a 2khz burst at full power. Let’s say with real music off by 50%… still close enough.

The rise itself is not the signal, the imbalance between low and high order harmonics is what signals that it is sounding too loud because those are unmasked and heard as unnatural. What you say about Keith’s experiment doesn’t make any sense…the added low order harmonics (mainly 2nd) provide the masking. It was clear from the listening results that 3rd alone does not confer the same benefit of masking to the other, unpleasant sounding odd harmonics.

You are not using efficiency correctly. Efficiency is the % of electrical energy converted into acoustical energy. Sensitivity = 112 + 10 log (efficiency). What you are calling efficiency is just another way to represent Sensitivity that is based on voltage and not power.

 

[DaveC]

Just to be clear by adding more drivers in parallel you increase sensitivity (since the load impedance is reduced while the drive voltage is unchanged) but efficiency is unaffected (since 1 watt is evenly distributed between all the drivers in the array). Assuming of course that sensitivity is 2.83volt/1 meter and efficiency is 1 watt/1 meter....

With tube amplifiers the efficiency spec is more useful since tube amps don't generally double power as the load impedance is halved.

Not true... you also increase efficiency. If you parallel 2 drivers with the same sensitivity you also double power but the net result is +6dB which requires 4x the power using a single driver, so the efficiency does indeed increase by 3dB.

 

[Atmasphere]

The rise itself is not the signal, the imbalance between low and high order harmonics is what signals that it is sounding too loud because those are unmasked and heard as unnatural. What you say about Keith’s experiment doesn’t make any sense…the added low order harmonics (mainly 2nd) provide the masking. It was clear from the listening results that 3rd alone does not confer the same benefit of masking to the other, unpleasant sounding odd harmonics.

You are not using efficiency correctly. Efficiency is the % of electrical energy converted into acoustical energy. Sensitivity = 112 + 10 log (efficiency). What you are calling efficiency is just another way to represent Sensitivity that is based on voltage and not power.

Efficiency is defined (as a specification) as 1 watt/1 meter. Sensitivity is defined as 2.83Volts/1 meter. But you aren't wrong:

Efficiency and sensitivity conversion - loudspeaker percent and dB per watt and meter loudspeaker efficiency versus sensitivity vs speaker sensitivity 1 watt = 2,83 volt box chart - sengpielaudio Sengpiel Berlin

Efficiency and sensitivity conversion - loudspeaker percent and dB per watt and meter dB/W/m loudspeaker efficiency versus sensitivity vs speaker sensitivity 1 watt = 2,83 volt box chart - sengpielaudio Eberhard Sengpiel

www.sengpielaudio.com

-except that in this case I was using efficiency as the defined specification- so I was using it correctly.

An easier way to look at this is sensitivity is based on voltage, since its fairly easy to get a solid state amp to behave as a voltage source; this spec took over in loudspeakers starting in the early 1980s.

Efficiency was used prior to that since tube amps do not always act as voltage sources; if they have no feedback at all they tend to act more like a power source. For more on this see:

Resources: Paradigms in Amplifier Design

Voltage Paradigm vs. Power Paradigm. Competing Paradigms in Amplifier and Loudspeaker Design, Test, and Measurement.

www.atma-sphere.com

SETs tend to be Power Paradigm devices and the speakers they drive are often designed with that in mind (although the designer may not have been thinking of the amp as a power source so much as he was thinking about getting the speaker to sound right with a given amplifier).

Not true... you also increase efficiency. If you parallel 2 drivers with the same sensitivity you also double power but the net result is +6dB which requires 4x the power using a single driver, so the efficiency does indeed increase by 3dB.

Sounds like that would violate Kirchoff's Law (the law of energy conservation). The part which seems problematic is the '+6dB' thing. First you cut the impedance in half, which requires 2x power rather than 1x power. That increases the sensitivity by 3dB since double the power is 3dB.

Where does the rest of the energy come from?

 

[DaveC]

Sounds like that would violate Kirchoff's Law (the law of energy conservation). The part which seems problematic is the '+6dB' thing. First you cut the impedance in half, which requires 2x power rather than 1x power. That increases the sensitivity by 3dB since double the power is 3dB.

Where does the rest of the energy come from?

The +6dB is not problematic, it's exactly how it works if you add a 2nd woofer in parallel. You halve the impedance so you double the power, I never said power stays the same. If you double the power into a driver you get +3dB out. If you have two drivers it results in +6dB. The extra +3dB comes from the increase in efficiency.

It's basically the same as doubling Sd or surface area of the driver. This increases the efficiency of transducing electrical energy to acoustic energy because the mechanical impedance match is better. It's a little bit similar to how a horn increases efficiency via matching impedance of driver to air.

 

[Atmasphere]

The +6dB is not problematic, it's exactly how it works if you add a 2nd woofer in parallel. You halve the impedance so you double the power, I never said power stays the same. If you double the power into a driver you get +3dB out. If you have two drivers it results in +6dB. The extra +3dB comes from the increase in efficiency.

OK- so try this on.

First you put 1 watt into a single 8 ohm driver. It makes 90 db at one meter.

Then you add a 2nd driver in parallel. The impedance is now halved. The voltage stays the same (2.83 volts), so now each driver is getting 1 watt.

The result is 93 dB.

From your text above it seems that you are assuming that both drivers would get 2 watts with this comment

If you double the power into a driver you get +3dB out. If you have two drivers it results in +6dB.

What am I missing?

 

[DaveC]

OK- so try this on.

First you put 1 watt into a single 8 ohm driver. It makes 90 db at one meter.

Then you add a 2nd driver in parallel. The impedance is now halved. The voltage stays the same (2.83 volts), so now each driver is getting 1 watt.

The result is 93 dB.

From your text above it seems that you are assuming that both drivers would get 2 watts with this comment


What am I missing?

No, I have no idea where you're getting that each woofer would get 2 watts.

You are incorrect that the result would be 93 dB, it would be 96 dB.

What you're missing is the effect of doubling Sd, or driver surface area. From my last post:

It's basically the same as doubling Sd or surface area of the driver. This increases the efficiency of transducing electrical energy to acoustic energy because the mechanical impedance match is better. It's a little bit similar to how a horn increases efficiency via matching impedance of driver to air.

 

[Solypsa]

I love it when what we assume tp be simple maths become (devolve) into a longer argument. No wonder this hobby is so convoluted

 

[advanced101]

I'm not even sure what we are arguing at this point. I think you Orchestral Classical guys have different measures of performance. In an effort to relate, I listened to Bartok tonight at full tilt. I'm with Solypsa, this hobby is extremely complex. If I put a new piece in and it sounds good, then that is that. It isn't a universal truth but a single, subjective opinion. We shouldn't have a debate as to why what I am hearing is technically flawed. The Manufacturers are not doing themselves a service trying to out nerd one another, and this is coming from a nerd.

 

[morricab]

Efficiency is defined (as a specification) as 1 watt/1 meter. Sensitivity is defined as 2.83Volts/1 meter. But you aren't wrong:

Efficiency and sensitivity conversion - loudspeaker percent and dB per watt and meter loudspeaker efficiency versus sensitivity vs speaker sensitivity 1 watt = 2,83 volt box chart - sengpielaudio Sengpiel Berlin

Efficiency and sensitivity conversion - loudspeaker percent and dB per watt and meter dB/W/m loudspeaker efficiency versus sensitivity vs speaker sensitivity 1 watt = 2,83 volt box chart - sengpielaudio Eberhard Sengpiel

www.sengpielaudio.com

-except that in this case I was using efficiency as the defined specification- so I was using it correctly.

An easier way to look at this is sensitivity is based on voltage, since its fairly easy to get a solid state amp to behave as a voltage source; this spec took over in loudspeakers starting in the early 1980s.

Efficiency was used prior to that since tube amps do not always act as voltage sources; if they have no feedback at all they tend to act more like a power source. For more on this see:

Resources: Paradigms in Amplifier Design

Voltage Paradigm vs. Power Paradigm. Competing Paradigms in Amplifier and Loudspeaker Design, Test, and Measurement.

www.atma-sphere.com

SETs tend to be Power Paradigm devices and the speakers they drive are often designed with that in mind (although the designer may not have been thinking of the amp as a power source so much as he was thinking about getting the speaker to sound right with a given amplifier).


Sounds like that would violate Kirchoff's Law (the law of energy conservation). The part which seems problematic is the '+6dB' thing. First you cut the impedance in half, which requires 2x power rather than 1x power. That increases the sensitivity by 3dB since double the power is 3dB.

Where does the rest of the energy come from?

Specification? What specification? You are simply using the word efficiency for a value that is a measure of sensitivity…full stop. The relationship to sensitivity is given by the equation I provided above.

 

[DaveC]

I'm not even sure what we are arguing at this point. I think you Orchestral Classical guys have different measures of performance. In an effort to relate, I listened to Bartok tonight at full tilt. I'm with Solypsa, this hobby is extremely complex. If I put a new piece in and it sounds good, then that is that. It isn't a universal truth but a single, subjective opinion. We shouldn't have a debate as to why what I am hearing is technically flawed. The Manufacturers are not doing themselves a service trying to out nerd one another, and this is coming from a nerd.

Sorry! My only comment was that power handling should be considered as well as efficiency. The rest is simply explaining how you can get a very efficient system that despite the high-eff spec may not be suitable for small amps and why that may be so.

I was hoping people might find it useful, I can't imagine how it's being interpreted as offensive... if it is, sorry that was NOT my intention!

 

[Duke LeJeune]

Sounds like that would violate Kirchoff's Law (the law of energy conservation). The part which seems problematic is the '+6dB' thing. First you cut the impedance in half, which requires 2x power rather than 1x power. That increases the sensitivity by 3dB since double the power is 3dB.

Where does the rest of the energy come from?

Excellent question! The following is my understanding, and I DO NOT claim it to be a complete understanding of the subject, but several paragraphs in we will encounter a factor which often gets ignored:

If we go from one driver to two, wired in parallel, assuming a voltage-paradigm amp, we get +3 dB from the doubling in wattage, and another +3 dB from the increase in efficiency (because we are moving twice the cone area), for a +6 dB gain in 2.83 volt sensitivity.

(If we go from one driver to two, with a power-paradigm amp, regardless of whether the drivers are wired in series or parallel, we get a +3 dB increase in efficiency.)

If we go from one driver to four and wire them in series-parallel, retaining the original net impedance, we get a +6 dB increase in both efficiency and voltage sensitivity, regardless of amplifier type.

BUT that CANNOT be the whole story! Consider this:

If the above trend is carried to the extreme, why couldn't we just keep adding drivers and end up with an efficiency greater than 100%? Obviously that can't happen, so there MUST be a missing piece to this puzzle!

The missing piece of the puzzle is the radiation pattern. The increase in efficiency from adding more powered cone area is accompanied by a REDUCTION in the radiation pattern coverage angle. To a first approximation, twice as many drivers = .707 times the net radiation pattern coverage area at a given frequency. And four times the number of drivers = 1/2 times the net radiation pattern coverage area at a given frequency.

So the increase in efficiency from adding drivers is accompanied by a decrease in radiation pattern coverage area, and in practice we can never squeeze enough drivers close enough together to violate the laws of physics.

It's a little bit similar to how a horn increases efficiency via matching impedance of driver to air.

Yes! And radiation pattern shows up here as well. There is a trade-off relationship between the horn's efficiency and its radiation pattern coverage area: The higher the efficiency, the smaller the radiation pattern coverage area (for a given driver).

 

[Solypsa]

This thread is becoming interesting to me as we dig into sensitivity / efficiency etc.

What effect does a passive crossover have ( if any ) on this equation? I see reference to 'complex crossovers' being an enemy of low powered amps. So if you have a two way speaker ( say a 15" woofer and a 1.4" CD tweeter for a common example ) with a 'full crossover' with coils on both woofer and CD maybe some additional correction ( more coils ? ) what does this mean in terms of energy available from amp?

 

[Duke LeJeune]

What effect does a passive crossover have ( if any ) on this equation? I see reference to 'complex crossovers' being an enemy of low powered amps. So if you have a two way speaker ( say a 15" woofer and a 1.4" CD tweeter for a common example ) with a 'full crossover' with coils on both woofer and CD maybe some additional correction ( more coils ? ) what does this mean in terms of energy available from amp?

The following is just my opinion.

It is nasty impedance curves and low efficiency which make life difficult for low-powered amps, rather than the crossover parts count in and of itself. The amplifier sees the load, not the number of crossover parts, and in some cases additional crossover parts make the impedance curve more benign. That being said, speakers with nasty impedance curves and low efficiencies do tend to have complex crossovers.

Passive crossovers filters remove energy, some of it deliberate (response smoothing and/or level matching), and some not (insertion losses are inevitable but are reduced with high quality parts). In the example you mention, 15" woofer + 1.4" compression driver, imo at a minimum the compression driver would need to be protected from bass signals, but in doing so you would (normally) also be making the load that the amplifier sees more benign.

 

[DaveC]

Excellent question! The following is my understanding, and I DO NOT claim it to be a complete understanding of the subject, but several paragraphs in we will encounter a factor which often gets ignored:

If we go from one driver to two, wired in parallel, assuming a voltage-paradigm amp, we get +3 dB from the doubling in wattage, and another +3 dB from the increase in efficiency (because we are moving twice the cone area), for a +6 dB gain in 2.83 volt sensitivity.

(If we go from one driver to two, with a power-paradigm amp, regardless of whether the drivers are wired in series or parallel, we get a +3 dB increase in efficiency.)

If we go from one driver to four and wire them in series-parallel, retaining the original net impedance, we get a +6 dB increase in both efficiency and voltage sensitivity, regardless of amplifier type.

BUT that CANNOT be the whole story! Consider this:

If the above trend is carried to the extreme, why couldn't we just keep adding drivers and end up with an efficiency greater than 100%? Obviously that can't happen, so there MUST be a missing piece to this puzzle!

The missing piece of the puzzle is the radiation pattern. The increase in efficiency from adding more powered cone area is accompanied by a REDUCTION in the radiation pattern coverage angle. To a first approximation, twice as many drivers = .707 times the net radiation pattern coverage area at a given frequency. And four times the number of drivers = 1/2 times the net radiation pattern coverage area at a given frequency.

So the increase in efficiency from adding drivers is accompanied by a decrease in radiation pattern coverage area, and in practice we can never squeeze enough drivers close enough together to violate the laws of physics.

Agreed, but it should be noted that it's frequency dependent... the higher the frequency the fewer additional drivers or surface area can be added before you cause beaming, or an overly narrow radiation pattern... The multiple drivers will act similar to a single larger drivers, so the radiation pattern will narrow according to the diameter of the equivalent radiating surface.

Also, at frequencies under the transition frequency of the room you don't have radiation patterns anymore and efficiency would depend almost entirely on the location of the drivers, whether they cause constructive or destructive interference, and it may vary depending on location within the room as well. But as far as subjective bass quality, IME the more surface area the better.

 

[DaveC]

The following is just my opinion.

It is nasty impedance curves and low efficiency which make life difficult for low-powered amps, rather than the crossover parts count in and of itself. The amplifier sees the load, not the number of crossover parts, and in some cases additional crossover parts make the impedance curve more benign. That being said, speakers with nasty impedance curves and low efficiencies do tend to have complex crossovers.

Passive crossovers filters remove energy, some of it deliberate (response smoothing and/or level matching), and some not (insertion losses are inevitable but are reduced with high quality parts). In the example you mention, 15" woofer + 1.4" compression driver, imo at a minimum the compression driver would need to be protected from bass signals, but in doing so you would (normally) also be making the load that the amplifier sees more benign.

Agreed... the ideal result is a resistive load, but many speakers don't compensate and you end up with ugly phase angles and unpredictable results when trying to match amps and speaker cables to speakers.

I'm now of the opinion that the ideal connection between the amp and speaker would include a network that would have to be custom made for every speaker and every type and length of cable because of this. It might not be more complex than a power factor correction network that varies with frequency.

So when people say the ideal speaker cable is low inductance, that's just an idealized theory, the truth is it really depends on the speaker's crossover. It makes selling speaker cables difficult, results are not predictable vs other types of cables. This is also why there are so many diametrically opposed designs for speaker cables with the manufacturers claiming their design is best. Obviously, someone is wrong, but it turns out it's all of them.

 

[Solypsa]

So maybe the cable should be ordered with the speaker to match it all?

 

[Solypsa]

@DaveC Iow what are the factors that a crossover design presents to you in cable selection?

 

[Atmasphere]

No, I have no idea where you're getting that each woofer would get 2 watts.

You are incorrect that the result would be 93 dB, it would be 96 dB.

What you're missing is the effect of doubling Sd, or driver surface area. From my last post:

It's basically the same as doubling Sd or surface area of the driver. This increases the efficiency of transducing electrical energy to acoustic energy because the mechanical impedance match is better. It's a little bit similar to how a horn increases efficiency via matching impedance of driver to air.

I think we've had a semantic issue: When I use the word 'efficiency' when it relates to loudspeakers, its always in the context of the efficiency specification, which is defined as 1 watt/1 meter.

So in this context if you double the number of drivers in an array that is getting one watt, each driver gets half as much power. The increase in efficiency seems to have more to do with lower frequencies, where you need more cone area to be able to support the waveform. But anecdotally, I've not seen the 'add another driver, get greater efficiency' thing play out; my speakers at home (Classic Audio Loudspeakers model T3) started life with a single 15" TAD woofer; a few years later a second woofer was added and even though both woofers operate in the same range, the overall efficiency spec did not change; they are still about 98dB 1 watt, 1 meter (actually more like 97.5dB for those being picky). I assume at higher frequencies you get cancellation issues that prevent the efficiency thing playing out. I say this as a local customer has been experimenting with speakers and has made systems of 1, 2, 3 and 6 drivers all of the same type (full range) in a single cabinet and they all measured the same. He has a garage full of "planters" (as he calls them) from all the empty cabinets he made. His cabinets were open back, and rolled off about 150Hz so it seems that the cabinet was preventing any efficiency benefit from the additional drivers insofar as low frequencies were concerned.

Linkwitz does provide a proof for your position, but he is careful to point out that the 6dB thing assumes a voltage source (otherwise if 1 watt is applied instead of 2.83 volts you get a theoretical 3db increase). He is also careful to point out the relationship between the cone diameter and the wavelength, suggesting that the benefit is greater for lower frequencies with respect to the particular driver (also assuming that the cabinet supports the additional required volume). Since we are dealing with zero feedback tube amplifiers as the core of this thread topic, I have been assuming that it was understood that the amplifier in the context of this thread was going to be constant power rather than constant voltage since a zero feedback tube amplifier was involved. The efficiency spec is far more useful when a tube amplifier is involved, since that spec does not assume constant voltage by definition. Put another way, tube amps don't double power as the load impedance is halved like a solid state amp can.