Talk:Tube sound/Archive 1

This page is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

I don't believe this has anything to do with triode characteristics. Valve amplifiers differ from transistor ones in a number of fundamental ways. The most obvious is the need for a final output transformer in valve systems which is not present in transistor designs. More subtle influences are the behaviour of valve circuits when handling signals outside the designed peformance where a transistor has a charactristic clipping effect whilst the valve is just non-linear.

An output transformer is not required in e.g. circlotron designs with multiple paralelled 6C33C valves; a single pair has 20 ohm output impedance without feedback, and 500mA DC current capability (about 2A peak, IIRC). Similarly, transformer coupled MOSFET and power JFET designs have been implemented. While the use of a transformer certainly contributes to the characteristics of typical valve amplifiers, it is not necessarily defining to the positive aspects of the sound.

As someone once stated, I like valve amplifiers despite their distortion, not because of it. This article should reflect that the valve enthusiasts are divided between those who like the "valve sound" as they call it, and those for whom valves are simply a better means to an end in some cases. You should have a look at the transfer curves at some point. Valves typically have a more euphonic harmonic overtone spectrum than BJT and MOSFET devices before feedback is applied.

Zuiram 05:28, 3 November 2006 (UTC)

In any event, triodes were not the dominant part of the amplifier as to handle the necessary power to drive the transformer it was necessary to use a pentode or pentode pair. Triodes do make up a good part of a pre-amplifier but audio buffs seem to prefer transistorised preamplifiers in conjunction with valved power amplifiers! Rjstott

Claiming that it is necessary to use pentodes to achieve sufficient output power makes no sense. The 6C33C glass triode can deliver about 10-15W single-ended or 60W push-pull class A, while the 3CX300A1 ceramic triode can deliver a few hundred watts push-pull class A. GM70 also springs to mind.

It is easier to make good solid state preamplifiers than to make good solid state power amplifiers, amongst other things because of higher availability of parts (2SK389/2SJ109 dual monolithic JFETs come to mind), and the ability to run things with a very high bias without needing a lot of heat sinking. That said, I've not noticed that particular trend. I have, however, noticed that many valve enthusiasts like passive, stepped-transformer based volume controls as "preamps".

Zuiram 05:28, 3 November 2006 (UTC)

I agree. Device transfer characterstic is not as important as circuit topology. Pentodes are very similar to FETs, so some audiophiles mistakenly assumed you could produce "tube sound" with FETs. There was a great article in the early 1990s in Electronic Design or EDN or something about this. The author had designed amplifers for Carver (Carvin?). He had successfuly reproduced tube sound with bipolar junction transistors. I remember calling him directly to talk about it. It was really interesting. I've been looking for that article ever since. Anyone know what I'm talking about?? Madhu 21:16, 8 November 2005 (UTC)

There is no doubt that the topology is important. But the device transfer characteristics should not be ignored either. Pentodes are indeed somewhat similar to MOSFETs, which is what I imagine you meant. Triodes, however, pretty much define the valve enthusiasts who want neutral sound, and until recently there were no reasonably priced solid-state equivalents of the triode.

At a higher price, you can have Sony VFETs (200V 5A 160W JFETs) that have even better linearity than triodes with the same general transfer characteristic. If you can live with a single polarity, such as when using balanced circuits, SRPP, followers or transformer coupling, then Lovoltech has a series of power JFETs designed for laptop power supplies (24V 100A 30W). The importance of the device transfer characteristic is easily demonstrated by using a simple circuit, such as the Son of Zen by Nelson Pass, and building it with different devices.

In short, the result is the sum of the parts, and device transfer characteristics, topology and power supply design are generally considered equally important by many notables, such as John Curl (Parasound, etc.) and Nelson Pass (Pass Labs, etc.).

Geeze, I'm sticking in

I'm pretty sure this isn't true. I think it just has soft

audio level compression. - Omegatron

15:47, Feb 6, 2005 (UTC)

If it is true, it probably has more to do with a weak, unregulated power supply than anything else. Madhu 21:16, 8 November 2005 (UTC)

I think it is true. Its called gain compression in RF amps and such. Anyway whats the difference between soft limiting and gain compression?--Light current 15:08, 31 January 2006 (UTC)

"Gain compression" in RF amps is just non-linear distortion, or soft clipping. Here's a decent description: [1]

I would imagine tube audio amps don't do anything different. The only way I can think of that they could is if there were some element of the tube that was heated by higher signal levels and the resistance change from the heat also changed the gain?

I don't know enough about tubes or the circuits they live in. I thought only the filament got appreciably warm, though, and that isn't affected by the signal and doesn't affect the gain, as far as I know...

Oh, and the difference between

audio level compression

is that

• in clipping, the signal is distorted to keep it under a certain level, creating extra harmonics. Soft clipping just means there isn't a sharp "knee point" in the transfer characteristic, as the above gain compression article explains. A sine wave going through soft clipping would become more like a "smooth" square wave and have lots of extra harmonics.
• in compression, the gain of the circuit is actually changed in response to the level of the input over time, so the transfer function is linear as long as you're only looking at a short period of time. A sine wave in will look like a sine wave out, but the overall gain varies depending on the level of that sine wave. (Above a certain level, the output sine wave will always be the same amplitude.)
• "limiting" can mean either, apparently, so I crossed that out in my comment above and replaced it with clipping, which is the definition I meant. — Omegatron 15:35, 31 January 2006 (UTC)

Yes well i think we're in broad agreement on this. Its largely a matter of terminology. But to my knowledge, its the actual non linear characteristics of the device when run at high levels that causes the compression of gain. It happens in transistor RF amps too. I dont think its to do with heat altho' I could be wrong on that.--Light current 15:42, 31 January 2006 (UTC)

Simple enough.. At low currents, the slope is not very steep, at medium currents, the slope is linear and steep. At high currents, however, something interesting happens, which is not due to the valve itself. The output transformer has a significant AC impedance, and at high currents, the voltage drop over the output transformer becomes very significant, which lowers the plate voltage. At lower plate voltages, less current conducts.

Hence, it is the interaction between the transformer and the valve that give rise to this "soft" distortion characteristic.

I have an electrostatic headphone / valve driver combo, which does not use an output transformer (since the voltages involved are highish (+/- 350V IIRC) and the current is low (<10mA)). It does not have a soft clipping characteristic, since there is no output transformer.

Zuiram 05:28, 3 November 2006 (UTC)

Go read the article. It explains that "gain compression" is just non-linear distortion . — Omegatron 15:44, 31 January 2006 (UTC)

Yes its effect must be one of distortion, due to the nonlinearity of the transfer function which also, of course causes a loss of

Ahhh. Yes, it causes loss of signal level, but that's not "loss of gain". The gain of the circuit is still the same; it's just reached the max it can output for a given input. — Omegatron 16:06, 2 February 2006 (UTC)

I meant loss of 'slope' or 'differential' gain. ie we have gain compression. I use the term 'slope' as you would use it in 'slope resistance' for a diode etc.--Light current 17:10, 3 February 2006 (UTC)

Your argument is correct, however, many data sheets for RF amplifiers list gain compression rather than non-linear distortion simply because it's easier to measure. Further, in narrowband systems, the effect "looks" more like gain compression simply because the harmonics are filtered out. In wideband and low frequency systems, the non-linear effects are readily visible, e.g. the output is clipped. To see the same thing at 1 GHz, you would need an oscilloscope with a bandwidth of at least 10+ GHz. Yes, you could see it with a spectrum analyzer, but what you see is the fundamental compressed and the harmonics picking up. It's not as intuitive as time domain clipping. Madhu 18:33, 2 February 2006 (UTC)

Yeah, RF tube amps are from a very different world from audio tube amps. Is "gain compression" only relevant to AM? — Omegatron 19:18, 2 February 2006 (UTC)

I would say not. As you turn up the wick, with whatever sort of signal you have, you will get gain compression. Now maybe its the time for me to mention the fact that a transistor's operating point may move with temperature, so higher power o/p may lead to compression due to collector dissipation. Im not sure about a tubes operating point with output voltage. I would think it woudnt move much at all( theyre running damn hot as it is!!)--Light current 17:10, 3 February 2006 (UTC)

But it's not a change in gain; it's non-linear distortion. The output level stays relatively the same as the input level goes higher. Gain is a

linear operation. — Omegatron

17:19, 3 February 2006 (UTC)

Yes, the term "non-linear distortion" is a little redundant, since all distortion is non-linear and all non-linearities cause distortion.

I agree with everything you said here, but I don't know why it's called "gain compression". It's just a matter of semantics, I'm sure. — Omegatron 19:39, 3 February 2006 (UTC)

Tube or transistor, gain compression means the same thing and is caused by clipping, soft or hard. Here's an article about it. It's relevant in any system with a wide dynamic range, audio or RF. Front end RF amps are as susceptible as any. Years ago, we added a low noise RF amp and directional antenna to a consumer 900 MHz receiver. We hoped to improve the transmission range. It worked, but it also picked up a couple of UHF stations around 700 MHz. Turns out, channel 54 was blasting 6 MW (yes Megawatts) of AM, FM, and PM our way. Our poor little RF amp, expecting -80 dBm, was way out of it's league and splattered mixing products all over the place. Needless to say, there was considerable gain compression going on. Madhu 00:50, 3 February 2006 (UTC)

I've been meaning to illustrate the clipping and compression articles with some waveforms, as I tried to do on the compression talk page. I think I know how to illustrate them now. Maybe I'll do that tonight. — Omegatron 15:48, 31 January 2006 (UTC)

So I was thinking I could illustrate the different concepts by showing a simple waveform in both a "wide view" to show the change in gain over time, and a "zoomed in view" to show how the waveform is either distorted or not distorted. Here's the example wave I am thinking of, but maybe someone has a better idea:

It's a mixture of 100 and 110 Hz sine waves, to get the beating effect (so that it has some structure visible from this "macroscopic" view but still looks like a sine wave up close), then I increased it linearly from left to right, so you can see as the clipping/compression kicks in. I'm sure there's a better way to show this, though, so I'm holding off on making all of the images/sound files for a bit until I get some other opinions. — Omegatron 02:34, 2 February 2006 (UTC)

A

Yes, that's exactly how it works. — Omegatron 16:06, 2 February 2006 (UTC)

Im glad we agree on something!--Light current 18:55, 2 February 2006 (UTC)

Hooray! — Omegatron 19:18, 2 February 2006 (UTC)

This acticle does not mention production and playback equipment in recording studios and that many producers with 'Golden ears' still prefer to do all the production (with real analog valve sound), including mixing down, prior to digitsation.

That is becoming rather rare, unfortunately. There are some places that still do it, though, like that recording studio which used to be a church and now does direct-to-disc vinyl recordings with valve equipment. Zuiram 05:28, 3 November 2006 (UTC)

I've heard these claims that even-numbered harmonics are "more musical" than odd-numbered, but I haven't seen a clear explanation. If the notes of the harmonic series are transposed down to stay within the octave, don't they all map to named notes? — Omegatron 02:40, 24 January 2006 (UTC)

The odd harmonics starting at the 5th are more offensive than the even ordered ones and the 3rd.

Someone (BBC?) did some experiments at one point in time and determined a weighting formula for the harmonic spectrum that correlated with subjective listening experiences in double blind trials. You could look that up.

2nd and 3rd order harmonics are generally not offensive, and at the levels present in linear circuits (even when operated open-loop, just about any simple audio circuit not using op-amps or other super-high-gain elements will generate <1% THD) are not audible in general.

Double blind experiments indicate no audible differences in a digital-sourced reference system when 2nd harmonic distortion is added until somewhat above 1%. For comparison, the threshold for the 7th harmonic is on the order of parts per million in the same experiment. This is part of the reason why there is a trend toward lower global negative feedback, as open-loop linear circuits generate a monotonically falling spectrum of harmonic overtones that is consistently porportional to the signal level, which closely mimics the distortion in our ear.

Zuiram 05:28, 3 November 2006 (UTC)

Not necessarily. Harmonics 2, 4, 8, 16 etc. are by definition named notes. If one assumes the frequency of each half step is given by:

${\displaystyle A_{o}(2^{\frac {1}{12}})^{N}}$

where N is the step number (e.g. A is 0, A# is 2, B is 3, C is 4 etc) and Ao is the root note (440 Hz for A above middle C, I think). Every power of 2 is exactly an integer step, which is a named note. If you do the math, the 3rd harmonic is 19.02 half steps up. This one octave and 7.02 half steps. For example, the 3rd harmonic of C is between G and G#. The 5th harmonic is 27.863 half steps, or two octaves and 3.863 half steps. That's equivalent to C and something between D and D#. The 6th harmonic isn't great either, it's 31.02 half steps. 4th and 8th harmonics are precisely the same note one and two octaves up. Madhu 03:16, 26 January 2006 (UTC)

Hmm... I guess I was wrong.

I guess I never thought it through above the 6th.

In equal temperament, as you described, they're all going to be off, but the article said Just intonation. Still, though, in JI, the named notes are derived from the harmonic series, so I don't understand how any of the harmonics could be "unmusical". And, it seems, the offenders are really prime harmonics, so they're necessarily going to be odd. — Omegatron 05:22, 26 January 2006 (UTC)

2nd and 3rd are prime, and not generally offensive. A single-ended non-feedback valve amp, for example, will generate "exclusively" 2nd (from the valve) and 3rd (from the transformer) distortion. All dynamic loudspeakers generate significant (up to 30%) 2nd and 3rd harmonic distortion at normal listening levels. Push-pull type electrostatics generate only 3rd order (<1%).

Zuiram 05:28, 3 November 2006 (UTC)

This JI page considers 7/4 to be a minor seventh. And here's another:

The 7:4 ratio is the naturally occurring minor seventh that exists in the overtone series (approximately 31% of a semi-tone flat from the equal tempered minor seventh). [2] — Omegatron 19:40, 31 January 2006 (UTC)

Here's the root of the problem (pun intended): equal temperment results in irrational multpliers for notes. So by definition, it is not possible for any two notes (other than octaves) to be exact harmonics. We all know that irrational numbers cannot be written as the ratio of two integers, but in many cases, it's close enough (as your table shows). I think the problem with odd harmonics (and some even harmonics) is that they are too far from any "good" note. Based on this argument, I tend to think tuning a guitar using frets might be more accurate than tuning using harmonics. This assumes the frets are placed precisely. In reality it probably doesn't matter all that much. I heard many "goldean ears" suggest that electronic tuners are not accurate -- I think the opposite is true. They are too accurate! If you tune an instrument based on good quality sound, it probably is not in precise equal temperment. Just my $0.02 Madhu 16:36, 26 January 2006 (UTC)

Well, yes, that's Just intonation (pure harmonic relationships) vs equal temperament (equally spaced intervals), but regardless of which you pick, it seems the prime number harmonics above 7 don't map to any of the 12 standard musical notes (exactly or approximately).

Is this important to the sound of distortion? I don't know. — Omegatron 16:58, 26 January 2006 (UTC)

That's the real question. My guess is that many harmonics are not pleasing, but the even/odd harmonic issue is probably as much rumor as reality. I think it's safe to say that there are so many differences between tubes and transistor amps that it's not exactly clearcut. Madhu 03:14, 27 January 2006 (UTC)

The even/odd issue started out as a rule of thumb, and unfortunately stuck.

The difference in harmonic spectra is relevant, but many people who use the terms "valve sound" and "transistor sound" are basing their comparisons on stuff from both camps that is far from state of the art. They feature differences that primarily result from design philosophy differences and the constraints associated with each device. The actual intrinsic differences only become clear once you compare state of the art (by which I do not mean "most expensive") equipment from each camp.

Zuiram 05:28, 3 November 2006 (UTC)

I think you are correct 'O' about the 7th, 11th and 13th harmonics not mapping to any of the 12 standard musical tones. Hence the problem of tuning painos. On a stretched string, the seventh harmonic (more strictly 'partial') is, for some reason, out of tune with the others. I think its slightly flat and therefore the notes have to be tweaked up slightly. Its all a compromise tho'. Do we really need to consider what happens above the seventh harmonic in valve sound?--Light current 23:52, 28 January 2006 (UTC)

In single ended triode, you can disregard the seventh, as it simply doesn't occur at any relevant levels. In push-pull pentode, at the other end of the spectrum, you'll see a fair bit of the seventh. Zuiram 05:28, 3 November 2006 (UTC)

Even your tender ears wont hear the 7th harmonic of 3kHz!--Light current 23:53, 28 January 2006 (UTC)

Well, you may hear the consequences of ultrasonic signals if it causes IMD or slew rate distortion in the amp, or even fries your tweaters ;-) I agree that 7th from a tube will normally be at a very low level - but my point is that its oversimplistic to believe that you can ignore anything outsde the audio band when designing an audio amplifier. I have experienced Mosfet power amps that actually (unintentionally) broadcast the music being played through them on FM, swamping BBC radio 2 VHF for several miles. Elsewhere you discuss compression. Powere from oscillations miles outside the audio band can compress your circuit just as easily as the stuff you can hear tubenutdave 02:26, 3 January 2007 (UTC)

The 7th harmonic of 3kHz is 21kHz, which is outside the hearing range of the general population due to incremental hearing loss with age. It is, however, not inaudible to fairly young listeners that have taken good care of their ears.

More relevantly, you'll definitely hear the 7th harmonic of 30Hz, which is 210Hz, at which point your ears are about 20dB more sensitive, meaning that even 0.1% 7th harmonic will sound like 10%. And, considering that about 50% of the audio power is in the band below 315Hz, it is easy to see how higher order distortion can "stick out" higher up.

Our ears are mainly phase sensitive below 1250Hz, meaning the low order distortion does not generally impact the sense of realism as much in that range. They are, however, mainly level sensitive above 1250Hz, meaning that the high order overtones of the high-power range end up in the most level-sensitive part of our hearing. Most valve designs avoid this.

Zuiram 05:28, 3 November 2006 (UTC)

No, the harmonics are perfect by definition. They are always in tune with each other because they're a perfect mathematical relationship. That's what I don't understand about them being "unmusical". I don't think it's possible for perfect harmonics to be "unmusical", since

musical intervals

are based on the harmonic series.

beat frequency

at 5 Hz, and are quite dissonant. A 100 Hz and its 7th harmonic, 700 Hz, would cycle at 100 Hz. I don't know how you can be more consonant than a harmonic.

As for pianos, the strings are not ideal (they are stiff, three dimensional, elastic, and so on), so the harmonics generated are not perfect. All of the harmonics on a real stringed instrument are slightly sharp from the ideal mathematical harmonics. (They become "

inharmonic partials

", though they are usually just considered harmonics since they're "close enough".) There's no analogous mechanism for harmonic distortion, so all the harmonics are exact multiples of their fundamental.

Yes, we do need to consider what happens to the harmonics in valve sound, simply because valve enthusiasts use it in their arguments. Besides, you can hear the 7th harmonics of every sound below ~3 kHz. — Omegatron 02:21, 29 January 2006 (UTC)

There are several problems here that we should take care not to confuse. The differential topology tends to cancel even order harmonics (regardless of if a tube or transistor amplifier) Even order harmomnies tend to sound like rich fat chords. Odd order harmonies my still be exact numerical ratio's but even a perfectly tuned piano can sound bad when someone hits a wrong note, a sharp or flat that really shouldnt be there (where is Les Dawson when you need him to demonstrate ;-) ) The real problem is distortion products that are not harmonic at all, which tend to come from real world component "variances from the ideal" (notably dialectric absorption and ESL/ESL by capacitors etc) and NFB, both of which can "smear energy in the time domain". Real world amplifiers do not have perfect impulse response. (also because that would require infinite slew rate and bandwidth). non harmonic distortion products do ot sound nice.tubenutdave 02:26, 3 January 2007 (UTC)

Most Perfect harmonics wont be unmusical, but in the equal tempered tuning system, most of the notes are inaccurate anyway so their harmonics sound bad against other notes in the eqaual tempered scale.See table below. Also, as said above I think some instruments dont produce perfect harmonics when you go higher up.

If the first 15 harmonics are

semitones, and in relation to that scale, many of the harmonics are slightly out of tune, and the 7th, 11th, and 13th harmonics are significantly so. In the late 1930s, composer Paul Hindemith ranked musical intervals according to their relative dissonance

based on these and similar harmonic relationships.

Below is a comparison between the first 20 harmonics and their equivalent frequencies in the 12-tone equal-tempered scale. Orange-tinted fields highlight differences greater than 5 cents, which is the "

just noticeable difference

" for the human ear. (Because physical characteristics of musical instruments cause significant variations from these theoretical values, they should not be used for tuning without adjusting for those variations.)

---

Harmonic Note Variance 1st C1 0 cents 2nd C2 0 cents 3rd G2 +2 cents 4th C3 0 cents 5th E3 −14 cents 6th G3 +2 cents 7th Bb3 −31 cents

Harmonic Note Variance 8th C4 0 cents 9th D4 +4 cents 10th E4 −14 cents 11th F4 −49 cents 12th G4 +2 cents 13th A4 +41 cents 14th Bb4 −31 cents

Harmonic Note Variance 15th B4 −12 cents 16th C5 0 cents 17th C#5 +5 cents 18th D5 +4 cents 19thbgcolor=#eeeeee | D#5 −2 cents 20th E5 −14 cents

--Light current 02:45, 29 January 2006 (UTC)

What does equal temperament have to do with anything? The article is talking about Just intonation and harmonic distortion. — Omegatron 03:01, 31 January 2006 (UTC)

So you wanna talk JI eh? OK. Quote from article:

It is interesting to note that in the harmonic series even harmonics always tend to correspond to a named note (in Just Intonation) relative to the fundamental, whereas (SOME) odd harmonics do not (take for example the 7th, 11th and 13th harmonics).

So by your own table above, the 7th, 11th, and 13th harmonics are not musical 'cos they dont correspond to any of the other notes in the scale (major, minor,pentatonic or whatever). So the sentence should be modifed as above to be correct. Understand?--Light current 04:10, 31 January 2006 (UTC)

It is interesting to note that the 7th harmonic (7/4 ratio or 14/8) is nearly (but not quite) equivalent to a major 7th (15/8). It will of course sound flat. So it would appear I was right before -- strange but true! --Light current 04:36, 31 January 2006 (UTC)

Just found this on the web:

For a note C, an octave below middle C, the first 10 harmonics in the chord of nature would be: (image of stave with some notes appeared here) The accidental on the note B is a contemporary '3/4 flat' sign. This is included because the 7th harmonic is rather flat, making the chord of the first 8 harmonics a dominant seventh with an 'out of tune' seventh. When sound comes from a source that is not perfectly periodic, the chord of nature may still be inherent in the sound, but the chord will be slightly 'imperfect'. Musical strings are sources whose motion is generally very close to being periodic, so the chord of nature is clearly recognisable as inherent in their tone. If a sound source is not perfectly periodic, the 'pure tones' are then not true harmonics with frequencies arranged in the harmonic series, but may nevertheless be very close to this. Rather than 'harmonics', they are properly called partials. Any musical tone with a definite pitch can be 'diffracted' into partials, usually arranged in the Chord of Nature, much as white light can be diffracted into colours of the rainbow. The component simple tones within the Chord are beautiful to behold, and stand at the threshold of the inner world of sound.

Although the 7th harmonic or 7th partial is 'out of tune' by the standards of Western tonality, there is nothing 'wrong' with it. It is, in fact, absolutely perfect in its relationship to the other partial or harmonic tones - but to hear this one has to be free of the expectations brought about by Western musical conditioning.

So maybe I was right after all! --Light current 04:55, 31 January 2006 (UTC)

Yes. That's what I just said. — Omegatron 15:18, 31 January 2006 (UTC)

Yeah, but from a western musicians POV, the 7th harmonic will sound out. Which is why, as I mentioned earlier, piano tuners may have to tweak some notes just a teeny bit sharper to get the right sound. After all, its the highest partial that grabs most of our attention in a note (assuming you can hear it!)--Light current 15:49, 31 January 2006 (UTC)

Theres something still not quite right here. On one page (WP) I read that the 7/4 ratio is sort of half way between a

A large portion of this article seems to have been copied en bloc from [3]. I shall remove it all--Light current 03:55, 1 February 2006 (UTC)

I have recently been reading about the gain compression that takes place due to loudspaeker voice coils heating up and incresasing their resistance. This causes less power to be drawn from the amplifier and a reduction in SPL. Maybe we need a separate article purely on Gain compression. Any thoughts?--Light current 00:23, 2 Feb 2006 (UTC)

The voice coil heating phenomenon is well-known in pro audio circles, but not very well known in hi-fi circles. JBL 2226, for example, a 12" pro woofer, is rated at 3dB power compression at 600W, which is an exceptionally good figure. Most home audio drivers will have more than 3dB power compression at one tenth of that.

Zuiram 05:28, 3 November 2006 (UTC)

Does that have any real effect? As we've mentioned elsewhere, the output impedance of a loudspeaker amplifier is made very low, a bridging connection, so that the amplifier can counteract any such effects.

Yes, gain compression obviously needs an article. — Omegatron 19:39, 3 February 2006 (UTC)

The output impedance of the amplifier should be high to counteract that. The motion of the cone is proportional to the current through the voice coil. As the resistance of the voice coil has a positive temperature coefficient, the heating causes the resistance to increase, causing less current to be drawn for a given voltage.

Yes, a voltage driven loudspeaker (most of the ones out there) will need a low impedance source (or compensation of some sort) to control the system resonance. That is the reason for voltage-driven loudspeakers (rather than current-driven) and for voltage-output amplifiers (rather than current-output).

In fact, using current output and a biquad transform to handle the resonance, you generally reduce the distortion by approximately one order of magnitude for a given loudspeaker driver. I could point you in the direction of the research if you speak technical Russion. I don't, I just have the links and got an executive summary from someone.

Incidentally, Steen Duelund (RIP) from Denmark did some work on this, and actually used resistive elements with a negative temperature coefficient in his passive crossover networks to compensate for this effect. Apparently to great effect in terms of both micro- and macrodynamics.

Zuiram 05:28, 3 November 2006 (UTC)

Yes it has effect, not on distortion necessarily, but on overall efficiency of the sound system and so is a form of gain compression.--Light current 19:43, 3 February 2006 (UTC)

Compression is a form of distortion. One that is particularly offensive to those who enjoy dynamic music, such as Jazz. Yes, it also has an effect on the system efficiency, but valves are not exactly known for their high efficiency.

Zuiram 05:28, 3 November 2006 (UTC)

To be merged into this article by authorised audiophiles. --Light current 17:22, 5 March 2006 (UTC)