Jump to content

Talk:Tube sound/Archive 2

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
Archive 1Archive 2

'Audiophile amplifiers'

For diverse reasons, Although valves are today an obsolete technicnology except for specialist applications, there has been a resurgence in the popularity of valves for so-called "high end" audio amplification. There are doubtless aescetic and other supporting factors for this, but there is a strongly and widely held view that valve amplifiers simply sound.. preferable. (Audiofile electronics is a field where passions run high verging on religious, preclusing the use of terms such as "better"). The reasons why this is so are complex (for example being due to circuit topology, the transfer function of a valve compared to a transistor, and etc) and are heavily debated, nevertheless the effect is genuine, and is discussed at length in

In comparison with modern, primarily transistor amplifers, valve amplifiers tend to be rather low power, depending on the power tube used), and in particular and often low efficiency The "classical" valve amplifier uses the Directly Heated Single Ended Triode topology (DH-SET), a topology that uses teh gain device in class A. The typical valve using this topology in (rare) current commercial production is the 300B, typically yielding ~ 5 watts. It should be noted however that the simplicity of the DH-SET circuit lends itself to hobbyist construction, so an unknown but perhaps majority of DH-SETs in use today are unique constructions, albeit usually variations of a small number of basic designs. Many hobbyist constructors of audiofile amplifiers are (and are proud to consider themselves) extremists, and this is especially so for DH-SET constructors. A substantial minority of such constructors take minimalism and component selection to extremes, and many so called "flea powered" amplifiers in the 2 watt class are constructed, using tube types that became obsolete pre war, arguing that the minimalist designs have sonic benefits. However such low power output pushes complete replay system design problems firmly onto the louspeaker, requiring very high efficiencies, typically 10dB better than modern mainstream audio loudspeakers achieve, often horn speakers. A separate religious debate rages about the merits and demerits of different speaker technologies, but it is perhaps fair to say that whatever merits extremely low power DH-SET amplifiers may have are partly paid for by problems facing the loudspeaker.

A minority of home constructed DH-SETs use extreme tubes to yield up to 25 watts (or beyond) in class A, although the engineering considerations to achieve this are (notably constructing suitably massive coupling transformers while maintaining the desired wide bandwidth) as daunting. Class A amplifers are inherently very inefficient, so power supply and thermal considerations are also problementic for high power designs (which can be solved, but at a price, not purely financial.)

During the 1960's and 70's in particular (sometimes referred to as the "golden age" of valve amplifiers, (the height of thier development prior to the introduction of the transistor), the majority of commercial amplifiers adopted derivatives of the class B, push pull, negative feedback topology pioneered by Williamson, this yielding greater power and measured linearity despite using dramatically smaller (and cheaper) transformers. 12-20 watts were obtainable depending on the power tubes used (often EL84, KT66, EL34 or KT88), with high damping factors, and this facilitated the widespread introduction of relatively cheap to produce mutliway box speakers, and the "hi-fi" industry was born.

Signal amplifiers using tubes are capable of very high frequency response ranges - up to RF. Indeed many of the vavle types used in SET (Single Ended Triode) amplifiers are in fact designed to operate in the megahertz range in radio transmitters.

In practice however tube amplifier designs typically "couple" stages either capacitively or using transformers and these devices do limit the bandwidth at both high and low frequencies. Nevertheless audiofile power amplifiers have substantively flat frequency response accross and beyond the audio band, typically power amps having -3dB points order < 10 hz, > 65 khz. This contrasts with some transistor amplifiers that may go out far beyond 100 khz. Some specialist valve preamplifiers amplifiers (e.g. microphone amplifiers) may be essentially flat to beyond 100 kHz, and remain excellent and amplifiers of chocie (still widely used in studio's).

However, the majority of comercial audio preamplifiers made during the "golden age" have complex filter circuits for equalisation and tone adjustment. Today many consider that these devices are fit only for scrap, having both appalling sonics (see valve sound) and also appalling measured performance, notably frequency response and phase linearity. However there are always exceptions, and a small number of excellent commercial designs remain on the market to this day, from e.g. Audio Research and others. Yet the role of "preamplifier" is increasingly tied to teh minority of audiofiles still using vinyl, all "modern" sources beling line level and requiring only switching, buffering and volume control.

Output transformers

Do all valve amps need an o/p trans or is there some trick to directly connect an 8 ohm speaker say.?--Light current 18:14, 19 March 2006 (UTC)

If all valve amps need an o/p trans, then this will cause a major limitation on the BW and prevent the application of large amounts of NFB due to the nasty phase characteristics of transformers. --Light current 22:57, 30 March 2006 (UTC)

Try a Google search (or the search engine of your choice) for "OTL amplifier". --Reid 64.171.68.130 18:58, 18 October 2006 (UTC)
Valve amps generally do not apply large amounts of NFB, for that exact reason. Surprisingly, scaling down the NFB is not entirely negative. Note, though, that it is quite possible to achieve wide bandwidth (e.g. 10Hz-100KHz), but rather expensive. It also depends on the plate impedance of the output valve.
Actually the reasons for using limited amounts of excess gain and huge amopunts of NFB in a tube amp are more complex. the group delay / phase lags referred to above (also from interstage coupling capacitors) means that al designs with 3 or more stages (most power amps) are potetially oscillators if the nyquist criteria isn't given due regard. Quite aside from that tubes have limited voltage gain anyway, and a lot of voltage gain is lost in the stepdown of the OPT. Having to add yet another stage may be an unacceptable price for providing the extra gain needed to have more NFB. And as commented above, NFB isn't as perfect as many people seem to think, epecially when used for MUSIC (cf continuous sine waves.) An interesting think is that you cna add two sinewaves of the same frequency and you will get a sine, regardless of the phase of the components. Now recall the fourier transform. Sadly you cnat add up all those sine waves in the fourier tranfsform to generate the waveform you have in music without getting the phase right. And the phase lags of NFB get them wrong, and also knobble the transient response tubenutdave 02:49, 3 January 2007 (UTC)
There are many OTL (output-transformerless) amplifiers out there. You'll probably want to search for "circlotron", which is generally the best OTL topology. If you're looking to build one, you'll probably still want to consider an output transformer, though, since it does a nice job of avoiding 50-3500Vdc across your speaker terminal if one of the valves die.
Designing a transformer for e.g. 20 ohm to 8 ohm conversion is a lot easier to do, with good results, than the same for e.g. 6000 ohm to 8 ohm... A bifilar winding with primaries in series and secondaries in paralell will give you 32 ohm to 8 ohm, for example, with pretty near-perfect results.
Zuiram 05:28, 3 November 2006 (UTC)
Yes OTLs exist but they are very very rare and have al kinds of problems of thier own. A tendency to blow up and a tendency to blow up your speakers when they do are just two of many. ;-) very Hig Z out (very poor damping factor) is another. The need for HUGE capacitors in the output stage is another (capacitors are even less ideal than transformers) They are very interesting amplifiers, but they have never gained widespread acceptance and there are many godo reasons for that. They have merits. and also demerits. (as all other types of amp of course)tubenutdave 02:49, 3 January 2007 (UTC)

Audible differences

It seems rather strange that the previous vociferous utterances of the audiophile community have not yet been crystallised into a coherent view of what the audible differences are between valve and transistor amplifiers. Is this because they have difficulty in describing these subtle differences or because in fact there are no differences at all?--Light current 03:18, 21 March 2006 (UTC)

There is indeed a difference in sound. The problem is explaining this in words. We don't have many words to describe the characteristics of sounds, we have to use words like 'creamy' and 'crunchy' to describe tube harmonic distortion, which are words to describe texture. Does sound have texture? The 'texture' of sound can only be experienced by you, so you'll have to take the test yourself and see if there is a difference. Otherwise, you have to take everyone's word for it that tube amps sound much 'warmer' and 'creamier' than solid-states. Believe me, for the most part, they do, whatever that means.

Put it in the article! See if it can stand!--Light current 02:09, 25 March 2006 (UTC)

The typical characterization of the difference between a good neutral-sounding valve amp and a good neutral-sounding solid state (transistor) amp is that the valve amp is more organic, warm and smooth, with particularly good reproduction of the human voice and the recording venue's atmosphere. As the others stated, you'll just have to listen for yourself. Human synaesthesia is not uniform between individuals, so the words used to describe it will not mean the same to everyone, at least not until you have your own experience from which to form your own description.

That said, there has not yet crystallized a consensus on whether "valve sound" refers to the sound of good neutral-sounding valve amps, or to the "romantic" highly-colored valve amps. There is a major difference between those two approaches. Zuiram 05:28, 3 November 2006 (UTC)

Accuracy in recording studios

Dont all recording studios now use opamp and transistor gear for the very reason that valves cause much more distortion?--Light current 17:32, 7 April 2006 (UTC)

I'm sure there are a handful with all-tube fanatics. I'm certain that many of them use tube preamps or tube guitar amps because of their perceived value. If the singer thinks tube mic preamps sound better, you record with one. — Omegatron 18:47, 7 April 2006 (UTC)

I really dont know. I would think today's recording engineerrs decide on which kit to use.Im not, of course, talking about the bands own gear- that could be anything.

Im talking about the stuff the band dont bring with them! This will be chosen from the kit available in the studio. Most modern studios probable now use transistor kit. Therefore....blah blah.

Also I think any effects like 'warmness' (distortion) could be added at the mixing stage, but I dont know if thats how it happens. We need a real recording engineer to tell us how its done.

--Light current 19:08, 7 April 2006 (UTC)

This depends on what kind of recording studio you're talking about.
The good studios usually have a wide selection of equipment, and all commercially viable studios allow the artists a great deal of say over what goes.
But the selection of studios with valve equipment is dropping. I think this is mainly because the current artists aren't very quality-conscious (ref. modern <6dB dynamic range vs original 12-24dB), and because there are few manufacturers of valve studio equipment left.
When the Norwegian band Midnight Choir wanted to record their most recent album (Waiting for the Bricks to Fall), they had to go to Eastern Europe to find a studio that had the all-valve recording-mixing-mastering chain they were looking for. Of course, there are studios in the west that offer this, but they cost a fortune. (I think Sterling Sound Studios (G'n'R, etc.) has this kind of gear, for instance.) That album incidentally clearly demonstrates the more "classic" valve sound, in that the atmosphere, emotion, clarity and presence is outstanding, while the sound can not by any means be considered "neutral". One of the most enjoyable recordings I own, though. I just don't want my own stereo to add to or subtract from what they've already done.
As for "adding warmness", I haven't heard anyone succeed at that yet. You can, however, add a lot of second harmonic distortion (usually by rectifying the sound with germanium diodes, attenuating the rectified sound, and mixing it back in, or doing the digital equivalent). That will yield what is commonly referred to as a "phat" sound, which may be mistaken for warmth if you don't have grounds for comparison, ref. my earlier comments on synaesthesia. Zuiram 05:28, 3 November 2006 (UTC)

Supposed reasons for valve sound

Quote: "In fact, the harmonics produced by a non-linear device depend on the topology and symmetry of the amplifier; not the type of device used. An amplifier with a symmetric (odd symmetry) transfer function, like a solid state push-pull op-amp, produces only odd harmonics. An amplifier with an asymmetric transfer function, like a class A valve amplifier, produces both even and odd harmonics.[1][2] As valves are often run in class A, and semiconductor amplifiers are often push-pull, the types of distortion are incorrectly associated with the devices instead of the topology.

In order to produce only even harmonics, the device needs a transfer function with even symmetry. A simple example is a solid state full-wave rectifier. Note that the fundamental, which is an odd-numbered harmonic, would not be reproduced at all. (The lowest frequency produced by a full-wave rectifier is double the original; or the second harmonic.) The production of only even harmonics is obviously not desirable in audio reproduction systems, though it is used in guitar distortion.

In audio reproduction systems, the types of harmonics produced should be irrelevant, since proper amplifier design can reduce all harmonics to inaudibility, and they should never see overload conditions. It is, of course, possible that the greater amount of distortion in class A valve amplifiers is the actual reason for the perceptually "improved" sound, even if it is degradation from an engineering standpoint." End Quote

The harmonics produced by a non-linear device are dependent on the device used and not just the topology. The transfer function for a transistor and valve will be different, surely this is patently obvious. The quote that valves are often run in class A and semiconductor amps are push-pull is innacurate - push-pull AB is one of the, if not the, most common valve guitar amp output stage. I tend to agree with the stance on audio reproduction (Im a guitarist not an audiophile) but signal transients from analog input sources such as the magnetic pickups of an electric guitar or a microphone will cause significant distortion http://www.milbert.com/tstxt.htm . Shame the people that contribute to Wikipedia dont actually have to know anything...

Thats right. Even you could contribute- as you have! Please sign your posts tho' by typing 4 tildes ~~~~--Light current 13:46, 23 April 2006 (UTC)
So-called "audiophile" valve amps are usually single-ended triode or class A1 (no grid current) triode. "Audiophile" solid-state amps usually have more power, and therefore tend to use a lower conduction angle to avoid having to dissipate excessive heat. There are, however, a number of single-ended solid-state amps (e.g. Son of Zen (DIY), F1-F3 (First Watt), Aleph (Pass Labs)) and class A solid-state amps (e.g. Sugden) out there.
The choice of active device is clearly relevant, and is independant of the choice of conduction angle. Zuiram 05:28, 3 November 2006 (UTC)

Canadian English

Tended to use the term "Tube" -- Older tubes were sold by several brands under the trademark "Radiotron" (Marconi, GE, RCA) Rogers used the company name "rogers radio tubes" and I have at least one example where this was molded into the bottom of the base.

cmacd 12:35, 12 May 2006 (UTC)


Topologies and distortion

THe article says that types of distortion are more dependent on stage topology (configuration) than on the type of device (solid state or vacuum tube technology). 8-?

Would anyone like to justify this statement quoting distortion figures for common base, common emmitter, emitter follower and common cathode, common grid, cathode follower etc before I remove this claim? 8-|--Light current 22:31, 1 June 2006 (UTC)

I had no hand in writing this article, just a few minor link tweaks. However, I have "The Cool Sound of Tubes" in front of me and would like to browse the article to see if this statement can be supported (I am a subscriber to IEEE Spectrum). Gerry Ashton 22:45, 1 June 2006 (UTC)

OK Im eager to see if there are any legit claims. I think the question of distortion in any particular stage is going to be governed almost exclusively by the amount of local NFB applied. BTW I removed the non working link to that article "The Cool Sound of Tubes". 8-|--Light current 22:58, 1 June 2006 (UTC)

NFB is not all that relevant here. What you're looking for is the open-loop harmonic spectrum, which diverges significantly between triodes, tetrodes, pentodes, JFETs, MOSFETs and BJTs. While NFB is, in theory, perfect, in practice, applying excessive amounts of NFB ends up reducing the amplitude of less offensive harmonics at the expense of shifting them to a more offensive set.
Remember that the NFB has to be applied somewhere, and that somewhere is generally the input stage. In theory, again, that should correct things. However, in practice, the input stage is not linear itself, meaning that the correction is non-linear. Some of the papers describing this problem are old enough that you can't even find them online, so it's been known (and often ignored) for some time.
Most valve designers use a moderate-to-low amount of NFB, particularly because gain is much harder to come by in a valve circuit than in a transistor circuit. I've seen transistor audio circuits with 80dB of feedback and 40dB of closed-loop gain, while most valve circuits would be hard pressed to get to just 30dB of feedback. This also figures in the overall sonics.
Zuiram 05:28, 3 November 2006 (UTC)

Must have removed it from another article. Just had a quick read of that article. It doesnt have many hard facts, but it could be quoted from becasue, as you may know, WP is more interested in verifiability than truth!--Light current 23:33, 1 June 2006 (UTC)

I noticed your removal of "The Cool Sound of Tubes" link. Since that is one of my favorite Spectrum articles, I went to the IEEE web site and found where then had moved the article to. I then reinserted the link with the updated URL. BTW, I am quite sure that circuit topology can have a strong affect on the amount and kind of distortion that can occur, and strong negative feedback, either on a single stage or across several stages, can make the type of device seem irrelevant when tested with non-clipping sine wave inputs. The trick is trying to find a reliable reference that (a) will back this up at all, and (b) discuss what happens when playing real music rather than test tones. Gerry Ashton 00:01, 2 June 2006 (UTC)

Yes I agree! Over to you! 8-)--Light current 00:15, 2 June 2006 (UTC)

That raises another interesting point: clipping recovery. At any realistic listening volume, with typically not-so-efficient (<100dB/1W/1m) loudspeakers, both valve and solid-state amps clip regularly, but low NFB improves subjective clipping behaviour (I can dig up the link if you need it), and IIRC, valves recover from clipping a lot faster.

Also note that slew rate limiting can induce a clipping of sorts as well, and you'd be amazed at the HF stuff you end up finding inside the amp if you put a scope on it. Zuiram 05:28, 3 November 2006 (UTC)

Transfer characteristics of active devices

Perhaps Im being naive here, but I always thought that all active devices had non linear transfer characteristics: some square law, some exponential. Are there any that are truly linear (without NFB) 8-?--Light current 23:22, 1 June 2006 (UTC)

Define linear. Since all active devices, when operated for gain, convert voltage to current, there has to be some sort of input/output transfer function.

The most linear operation you can get, is with VFETs or triodes operated with significant source/cathode degeneration. While that constitutes NFB, it is not global NFB. The resulting I/V transfer function can be extremely linear. Zuiram 05:28, 3 November 2006 (UTC)

Plan to replace Supposed reasons for valve sound section

I am concerned about the Supposed reasons for valve sound section because it concentrates, at the beginning, on incorrect beliefs about why valve amps sound the way they do. I think it is more important to state valid reasons for why valve amps sound as they do. Eventually, we might address misconceptions, but I think we should find documented sources of those misconceptions.

Also, I feel the extended discussion of even-order distortion and odd-order distortion is not relevant. Although the discussion is partially correct (circuits with perfectly symmetrical transfer characteristics do not have even-order distortion) a peer-reviewed source (see my sandbox) claims that in practical audio amplifiers, the main difference in distortion is that valve amps tend to have lower-order distortion while semiconductor distortion has relatively more high-order distortion.

Mostly correct. Valve amps tend to have a monotonically falling harmonic spectrum which is monotonically proportional to signal level, whereas solid state amps tend to have a non-monotonic harmonic spectrum which is discontinously proportional to signal level. Zuiram 05:28, 3 November 2006 (UTC)

Therefore, I have created a section, Reasons for valve sound, in my sandbox. I don't have much hands-on experience with audio, and hardly any with valves, so I'd appreciate feedback from some of you who might be more knowledgeable before I make the change. Gerry Ashton 04:44, 2 June 2006 (UTC)

Even tho I dont agree with all (any?) of the stuff in the article you quote (cool sound of tubes), it is an independent source from a respectable journal and thus could be quoted in our article. I dont think that article would have been peer reviewed - but so what? It is verifiable and does not have to be true.8-(
However there are also 2 other refs already in the section and these views should be retained also. It would also be nice if pictures of the distortion spectra were refered to illustrating the diff betweeen semiconductor and tube dist.
It is best to write the section in a 'reported speech' style (eg. it is said that ...) rather than to assert 'facts' I feel. Basically every statement or assertion should be able to be referenced from a book, journal, paper etc. Personal views even based on ones own research/intuition are against WP policy. Hope you find this expo useful! 8-|--Light current 07:41, 2 June 2006 (UTC)
Ihave reread the article in question and actually found the diagrams. Now it appears to make more sense and appears to be well written material. Also I would have thought that an article of this nature would need to be peer reviewed.--Light current 19:04, 2 June 2006 (UTC)

To check on Light current's concern about peer review, I went to the IEEE web site and followed a chain of links from the Spectrum page to the information for authors page ([1]). This page states that IEEE publications are peer reviewed. I am not able to find any printed information in the 1998 issue about whether Spectrum was peer reviewed at that time, but I am a subscriber and never noticed any notice in the magazine about beginning peer review.

I'll think about switching to the reported speech style, although that seems stilted to me. I'd appreciate elaboration on what parts Light current disagrees with. I did take care to only report views contained in the article I cited, except for this statement:

Also, since valves are large and expensive, there is incentive to keep the open-loop gain of valve amplifiers low. In contrast, the cheapest and easiest way to build a solid state amplifier is to use an integrated circuit, which has extremely high open loop gain; there is no choice but to use massive amounts of negative feedback to reduce the gain to a practical level.

I consider that statement common knowledge, at least among electrical engineers.

When dealing with audio poweramps, which is where "valve sound" has any meaning, I'm not at all sure the assertion that valves are large and expensive have any real meaning. With the exception of transmitter valves, most valves have glass envelopes that are cooled by natural air convection. The solid-state devices of equivalent power, with equivalent conduction angle (operating class), are forced to use heat sinks to dissipate this excess energy.
Some valves (e.g. 6C45P, 6C33C) are perfectly happy having their carbon plates glowing a dull red from the heat, corresponding to temperatures in excess of 800 degrees Celsius, meaning they don't need much surface area to avoid overheating.
Solid-state devices generally do not handle temperatures in excess of 150 degrees Celsius, and to get any useful power handling out of them, you usually have to keep the temperature in the 50-75 degrees Celsius range. This requires a significant surface area.
Also, since the heatsinks are almost universally exposed in high-power non-pro designs, you effectively need to limit them to 55 degrees Celsius so people won't get burned when touching them. Valves also have the useful feature of glowing when they're hot (except when they're cooling down after use), so you know not to touch them.
While output transformers can be fairly expensive, a 50W quadfilar output transformer wired for 8 ohm to 2x64 ohm will not cost more than a heatsink that can service an equivalent (50W class A) solid-state amplifier.
Zuiram 05:28, 3 November 2006 (UTC)

Barbour's article contains distortion spectrum for 6 amplifiers and a transformer. I could scan some of them as fair use; any opinions on how many I can take as fair use? Unfortunately, the circuits are voltage amplifiers, not final power amps. I understand that the chief difference in tubes vs. transistors are thought to occur in the final stage. BTW, I have a 1959 RCA Receving Tube Manual in case there is anything from there that ought to be posted. Gerry Ashton 15:31, 2 June 2006 (UTC)

OK Gerry, i think the best course of action now would be for you to replace the sect in question with your version and let other editors see it/edit it some more. Please bear in mind that this can be a controversial subject and may attract a lot of 'interest'! I think the spectrum diagrams would each be worth a thousand words if you are sure they are fair use! 8-)--Light current 15:57, 2 June 2006 (UTC)
Eric Barbour is one of the contributors to WikiPedia. Perhaps he can be pursauded to release the diagrams himself.

cmacd 17:27, 2 June 2006 (UTC)

Ah well I dont know whether that would ceate a policy violation or not WP:NOR. The copyright now would be with the publishers IEEE now anyway. 8-|--Light current 17:55, 2 June 2006 (UTC)
That section was originally "Reasons for valve sound", and contained a bunch of myths about harmonic content, which are everywhere on the net. I tried to make it more accurate. I'd make some images of waveforms and spectra if desired.
Your sandbox looks good, though it says some of the same things as my section, and doesn't back up some of its other statements.
Putting up images you created does not violate policy. The copyright is not necessarily with the publisher, either. — Omegatron 18:32, 2 June 2006 (UTC)
Yes but 'O' you seem to miss my point which was that Eric Barbour having sold this article to IEEE spectrum probably does not now own the copyright to these images (or text for that matter)--Light current 19:00, 2 June 2006 (UTC)
Whoa. This article has become pretty cluttered since then. It's got a lot of redundancy.
Also, I don't see any reason why any of these sections should be replaced by the sandbox section. Just merge the sandbox into the article where appropriate, and clean up redundant parts. — Omegatron 18:35, 2 June 2006 (UTC)

I would be facinated to see supporting references for the information about harmonics in the exsiting section. I would also be interesed to know which parts of my version are perceived as not backed up, so I can clarify how it is backed up, or remove it. I have read the IEEE copyright agreement for its authors, and the IEEE does hold copyright, but it licenses back some rights to the author. Also, the IEEE does not pay authors. I also see that is considerable redundancy. Fixing that would require a major rewrite. I'll think about the redundancy issue.

I have created a new page, User talk:Gerry Ashton/collaborate where anyone who cares to do so can mark up my proposed replacement section, so their comments will be close to the text they are commenting on. Gerry Ashton 19:22, 2 June 2006 (UTC)

Thanks Gerry for clearing up those issues! BTW, when I said replace the section, I of course meant merge it with your own additions 8-)--Light current 19:24, 2 June 2006 (UTC)

Are odd harmonics 'musical'?

...or unmusical. That is: are they desirable or undesirable in an amp? Why 8-?--Light current 19:07, 2 June 2006 (UTC)

  1. All harmonics are bad in sound reproduction systems. You want as low distortion as possible
  2. Are some harmonics more musical than others? Well, that depends. Define "musical". — Omegatron 20:03, 2 June 2006 (UTC)

C'mon 'O'. We've had this discussion before. Can't remember what talk page it was on now. Common wisdom is that even harms sound OK whereas odd harms sound bad. Both of course are distortions that are not present in the original signal. Im innocently asking the question about odd harms becuase I dont know the answer. Musical means harmonious 8-|--Light current 23:09, 2 June 2006 (UTC)

Actually it was on this very talk page that we discussed harmonics and their musicality! How convenient! 8-)--Light current 23:23, 2 June 2006 (UTC)

I was just going to point that out. #Harmonic_series :-)
All harmonics are harmonious, aren't they? Which sound better to you? Shall I post some samples? Is there a scientific way to determine whether something sounds better or not? I've heard this notion many times before, and I'm skeptical as to whether it has any validity. Maybe it does, but I haven't seen any convincing explanation of why. — Omegatron 00:24, 3 June 2006 (UTC)
As I pointed out earlier here, there have been scientific studies regarding this. 2nd and 3rd are generally innocous. Higher order is detrimental, odd more so than even. A weighting factor is provided, and while I cannot remember it exactly, I do remember that the term was raised to the power of (N/4), where N is the order of the harmonic. Zuiram 05:28, 3 November 2006 (UTC)

Well as I said, this is the accepted wisdom. Im not convinced myself yet. However, I did find a web page earlier today that claimed to show the difference with sound samples containing even harmonics then odd harmonics. They did sound different, but I suspected some trickery or other messing with the sounds apart from harmonic additions. Ill try to find the page again and post it here. I would like to hear your samples. 8-)--Light current 00:42, 3 June 2006 (UTC)

They certainly sound different, but which is "more musical"? Which is better? I suspect it's purely subjective.
Ok, I'll upload something. Let me figure out how to post FLAC files, first.
How about this? Omegatron 01:11, 3 June 2006 (UTC)

OK thanks for that 'O'. Took me a while to find a codec for the .ogg files. Anyway, the tones:

  • first sounds pure (bit like a bass guitar with the tone turned down)
  • second sounds like an organ with a slightly breathy tone playing in multiple octaves - quite nice
  • third again sounds like an organ but not as pleasant as no 2 to my ears!

I would say the second is pref to the third. The first sounds perfectly ok to me being a bass player 8-)(I mean thats how it should sound -right?) --Light current 03:48, 3 June 2006 (UTC)

A bass would have all of those harmonics.  :-) — Omegatron 03:58, 3 June 2006 (UTC)

My bass guitar dont sound anything like those tones (only a bit like the pure tone). Each instrument will have different amounts of each harmonic but those tones had a lot of higher harms for a string bass! 8-|--Light current 04:13, 3 June 2006 (UTC)

Could you make some recordings? We don't seem to have any of bass guitar. — Omegatron 14:45, 3 June 2006 (UTC)

Maybe, if I get some time. There are so many options here that I think we need to say what we want!. What do you think would be suitable:

  • single notes,
  • over what pitch range
  • which strings (EADG) see bass guitar (BTW which bass would you like- I got both shown on this page!)
  • what length should each recorded sample be?
  • does it need to be enveloped or do you want the attack phase as well
  • any other requirements
  • I dont think I can make .ogg files. Whats the best alternative format?

Also I'll probably need to 'DI' into minidisc recorder then copy it into the computer and convert the file using Cool Edit before uploading. THis of course will give the raw sound without any cabinet effects. I think thatll be ok.

8-| --Light current 16:00, 3 June 2006 (UTC)

Any and all of the above.  :-) Whatever you have the time and energy to record is welcome. See Bassoon#Audio_examples, for instance. I don't see any bass guitar samples at all.
You have to use ogg format, but it's easy. For Cool Edit, unzip this file into C:\Program Files\Cooledit and then save as ogg. Otherwise, you can download dBPowerAmp music converter and the ogg codec and convert the file by right-clicking on it.
And if you do decide to upload stuff, put it on Commons, not here. — Omegatron 18:31, 3 June 2006 (UTC)

Actually, i may be able to record directly into the computer using Cool edit or some other package. 8-) any advice on how to do this will be welcomw. 8-|--Light current 18:52, 3 June 2006 (UTC)

Yes, that's what I meant. Record into Cool Edit and then save as an .ogg file using the above filter. — Omegatron 23:21, 3 June 2006 (UTC)
Keep in mind that the level of harmonics is also relevant. You want to stay at or below 1%, since even valve amps tend to do so at sensible listening levels. Zuiram 05:28, 3 November 2006 (UTC)

Facinating reference

I found a very interesting reference:

Russell Hamm, "Tubes Versus Transistors - Is There An Audible Difference?", Journal of the Audio Engineering Society, May 1973.

It can be found at [2]

It will take a while to digest this. Gerry Ashton 01:10, 3 June 2006 (UTC)

Good reference. This should definitely be in the article. — Omegatron 15:12, 3 June 2006 (UTC)
It already is! 8-) [3]. Just a clearer version here--Light current 16:04, 3 June 2006 (UTC)
I know; I just updated it. It's been there forever. I mean the information needs to be in the article. — Omegatron 02:06, 4 June 2006 (UTC)
Ahh! yes 8-)--Light current 02:30, 4 June 2006 (UTC)
This site seems to have some articles about valve sound from a legitimate perspective, too. — Omegatron 04:17, 4 June 2006 (UTC)
Yeah looks like a sensible fellow whose stuff can be referenced. I think maybe we need less exposition and more refs? 8-)--Light current 05:38, 4 June 2006 (UTC)

Three of the links in the article, in the External Links section, do not work for me:

Does everyone have trouble with these links? Gerry Ashton 00:29, 4 June 2006 (UTC)

Yeah. Maybe Milberts server is down!--Light current 01:49, 4 June 2006 (UTC)

Audible Differences

I question the article's claim under "Audible Differences" that "it appears that no results from scientifically conducted listening tests are available to confirm or deny the audiophile claims". I think these tests have been done and show that no detectable differences exist when the valve amplifier's distortion is low. For high distortion valve amps the difference is detectable, but only in the same sense as distortion is detectable. Without evidence I suppose this is just hearsay, but it is my recollection. Nowater57

Unfortunatley, we cant rely on peoples recollections, but if you can find a ref. then we can put it in--Light current 13:15, 5 June 2006 (UTC)
I also suspect these tests have been done many times. The aforementioned article (from 1973) says "Previous attempts to measure this difference have always assumed linear operation of the test amplifier. This conventional method of frequency response, distortion, and noise measurement has shown that no significant difference results." — Omegatron 13:30, 5 June 2006 (UTC)
Well of course we can reference the published Hamm article (but probably not the unpublished ones) But there doesnt seem to be hard evidence of controlled tests here. We must look elsewhere--Light current 14:17, 5 June 2006 (UTC)


Most Guitarists cant tell tube amps from SS

Read conclusion in this IEEE article from 1981 [4]--Light current 15:04, 5 June 2006 (UTC)

I searched on the IEEE web site and found the information needed to cite the article:
W. Bussey and R. Haigler (1981). "Tubes versus transistors in electric guitar amplifiers". IEEE International Conference on Acoustics, Speech, and Signal Processing. pp. Volume 6 p. 800–803. {{cite conference}}: Unknown parameter |booktitle= ignored (|book-title= suggested) (help) Also note that the summary is available at the IEEE website but the text is not, unless you subscribe to their electronic library service.Gerry Ashton 22:43, 5 June 2006 (UTC)

Well thats OK cos the text is available at [5] 8-)--Light current 00:32, 6 June 2006 (UTC)

Psychoacoustics

Good paper on ear 'self distortion' etc.--Light current 15:24, 5 June 2006 (UTC)

Shoot! forgot to include the link! Gotta find it again now! 8-(--Light current 00:44, 6 June 2006 (UTC)

Class A push-pull amplifier

Is there such a thing as a Class A push-pull amplifier, as mentioned in the article? —Bromskloss 20:25, 23 July 2006 (UTC)

It indicates here [6] that there is! 8-)--Light current 16:18, 24 July 2006 (UTC)
There is a tremendous amount of popular confusion on this subject, including the "Harmonic Content and Distortion" section of this page, which needs to be corrected. The author of that section is confusing "Class A" with "single-ended", which is very common (in no small part due to a compendium of inaccuracies and myth called The Tube Amp Book, by Aspen Pittman, which is tragically regarded as authoritative by a lot of people with no other background in electronics). The oppposite of "push-pull" is "single-ended". Either of these topologies can be class A, AB, B, or C, which describes how the gain devices (tubes or transistors) are biased (bias is sort of like the idle speed on a car engine). Operating class has nothing inherently to do with the circuit topology. It happens that in audio, all single-ended amps are also Class A, because they would sound horrible if they weren't, but push-pull amps can also be Class A if their gain devices are biased such that all devices conduct the signal at all times (i.e., they idle "hot" and no device ever falls into "cutoff"). Comparing "push-pull" with "class A" is a classic "apples & oranges" error. Class A also has nothing to do with whether the output is cathode-biased or fixed-biased, another common misconception promulgated by the book cited above. Amplifier operating classes are described well in the Wiki aticle on Electronic Amplifiers. --Reid (I should register...) 64.171.68.130 18:50, 18 October 2006 (UTC)


In practice, nearly all solid state amplifiers are push-pull, and they are almost invariably not class A, though many claim to be. Sugden has true class A solid-state amplifiers that are push-pull. Many valve amplifiers are push pull, and a large number of these are also class A.

The reason is quite simple: "true" class A effectively means that the idle current must be at least equal to the maximum signal current. This means tons of heat, which must be dissipated somehow. A valve can stand several hundred degrees celsius, while a transistor will need to be kept to a few tens of degrees above room temperature. That, in turn, means that it needs a lot of (expensive) heat sinking.

Simply put, the cost usually outweighs the benefit in solid-state amplifiers, so they are usually biased for "class A" operation up to a few watts (rarely more than the first 25W), which is really class AB. But you can find, or build, "true" class A solid-state amplifiers (360 degree conduction angle). The money will often be better spent elsewhere, though.

Preamplifiers are almost invariably true class A, whether single-ended or push-pull, since the relatively small power levels involved (5Vrms into a 75ohm load is about 300mW, and that's pretty much a worst case) don't usually necessitate any excessive heatsinking. Zuiram 05:28, 3 November 2006 (UTC)

Not long ago I tried to eliminate the entanglement of class and topology in the "Harmonic Content and Distortion" section. Then I found this comment on the discussion page. My advice is, if you see something wrong or unclear here, just dive in and fix it. Don't let misconceptions and twisted ways of looking at things persist. That section could still be better though. Among other things I left in the references to the external material which unfortunately doesn't really help clarify the issue, oh well.

And for whatever it's worth, the idle current in class A is at least half the maximum current -- the current varies up and down around the idle point. And an interesting thing to notice is that the power dissipated in the device is the voltage drop across the device times the current through the device, and the voltage drop and current are both varying with the signal, with the result that a device in class A that is driving a resistive load will dissipate maximum power (i.e., run the hottest) when there is no signal, and will run the coolest at maximum signal. Class B is opposite.

Atomota 06:20, 4 January 2007 (UTC)

well written answer, many thanks ! Its worth stating Light current (now banned) endlessly got hung up based on this apples and oranges confusion and a lot of his vandalisation was based on his misunderstanding of that
To (briefly) try to motivate why a designer may do BOTH, class A has many merits already stated, but used in its simplest SE form produces substantial (if monotonically decaying) distortion, and also modulates the supply rails directly with the signal. And in teh same way that NFB is imperfect, regulators (which are NFB circuits themselves) are imperfect and you can never clean up a power rail perfectly. If you now make (for example) a DIFFERENTIAL class A stage (aka a push pull stage), each "half" remains a class A as before, but additionally
  • the even order dist products tend to cancel, lowering overall distortion drammatically
  • PSRR is dramatically improved, as is CMRR
  • supply draw becomes (almost) constant, making regulation much simpler, (and the PSRR makes teh stage relatively insensitive to any ripple taht remains
from an engineering perspective push pull is a win win win approach - and doing it class A is the creme de la creme.As stated elsewhere howver it has one disdavantage. global warming ;-) tubenutdave 23:23, 9 May 2007 (UTC)

Removed text

I have removed the following text:

Class of operation

Similarly there is a tendency for mainstream commodity transistor amplifiers to operate in class AB1, hard towards class B and thus have significant crossover distortion during especially quiet passages of the music. whereas valve amplifiers are often pure class A (by definition is an SE amplifier) and even if class AB1, with a substantial class A region., so the amp is effectively operating in class A during quiet passages, with all the sonic advantages class A provides. But again it is possible to bias valves hard in to class B, and make class A transistor amps - and in such case the sonic qualities associated with this also occur

The statement that 'mainstream commodity' transistor amplifiers have significant crossover distortion is ludicrous IMHO. Even if the design is so bad as to have a dead band in the push-pull output stage, the generous amount of open loop gain combined with generous negative feedback reduces this notch distortion to insignificant levels. Note that I am not claiming that large open loop gain combined with a large feedback factor is 'good' from a sonic standpoint. I'm simply pointing out that in the 30 years or so that I have paid attention to audio amplifier reviews (both objective and subjective), I've not seen any evidence that crossover distortion is a common problem in mainstream solid-state amplifiers. Alfred Centauri 22:33, 8 April 2007 (UTC)

In what way "ludicrious". Pretty much every textbook that addresses class AB will discuss crossover distortion. The idea that arbitrarily large amounst of NFB will fix up all sins to "insignificant levels" is widely understood to be naive and the last 20 years or so amplifier design has pulled back from lousy open loop dist but huge open loop gain and resulting lage feedback margins in favour of producing better open loop linearity albeit with lower open loop gain and applying less feedback.
Statements like "operating in class A during quiet passages" are not simply unhelpful but actively misleading. By extension what you dont say is that as soon as you step out of a very quiet passage the amp enters class B and you get switching distortion - while teh signal is still low relative to full power. FACT is that true clas amps produce lowest distortion at zero power and distortion rises monotonically (and exponentially) as amplitude increases. Class B amplifiers produce lowest dist at near MAX power and dist increases as amplitude is REDUCED because teh switching dist is more or less constant and thus relatively larger as teh signal gets smaller. Crossover distorion is severe in the medium level region where and AB1 amp has just steped outside class A
More deeply, some (eg differentail) true class A amp circuits draw constant load from the PSU, making regulation easy and minimising interstage coupling via supply rail modulation. Class B amps generate varing load (includinbg being OFF) and thus invariably suffer from modulated supply rails and consequent interstage coupling, which often results in increased IMD and or more complex distortion spectra
It is generally accepted that class A is more linear than class B. The feedback argument is very old and tired. NFB has its virtues - but also its limitations tubenutdave 23:12, 9 May 2007 (UTC)

In what way ludicrous? Do you really believe that crossover distortion in mainstream commodity amplifiers is signficant??? Here's a task for you: Please provide the general transfer characteristic for a push-pull BJT emitter follower stage and then explain when and how this distortion is significant. If you need some help, please take a look at this: [7]

I disagree that "operating in class A during quiet passages" is misleading. Here's a question for you: What is a true class A circuit? Consider this: Imagine a class A push-pull amplifier rated at 10W into 8 ohms. Is it class A into 4 ohms? 2 ohms? Will this class A amplifier be class A at all frequencies into a complex speaker load? Consider the same amplifier except that, instead of clipping @ 10W into 8 ohms, it instead (and roughly speaking), slides into class B. Which will sound better - clipping or sliding into class B? In other words, would a 10W class A push-pull amplifier necessarily sound better than a similar 100W class AB amplifier with the same idle current? From what you've written, I believe you would answer with an emphatic NO. Yet, it seems clear to me (and any reasonable engineer) that, as long as both amplifiers operate within their limits, the amplifiers will sound similar. Yet, at a listening level where the 10W amp saturates on peaks, the 100W class AB amp will sound better (assuming it is not saturating too and assuming you aren't actually fond of the sound of clipping).

Further, the statement that "Crossover distorion is severe in the medium level region where and AB1 amp has just steped outside class A" is just plain wrong or at best, confused. Such a statement reveals an utter ignorance of the subject. Once again, peruse the paper I linked to above and then explain to me how you can justify such a statement.

Finally, you said "It is generally accepted that class A is more linear than class B." Well, I would say that is an understatement as is this "NFB has its virtues - but also its limitations." Any competent EE understands this intuitively. But this: "The feedback argument is very old and tired" is spoken like a true non-expert in the field. Alfred Centauri 02:12, 11 May 2007 (UTC)

Monotonic decay

Triodes (and Mosfets) produce a monotonically decaying harmonic distortion spectrum.

(1) How many lay readers have any idea what that means? (2) Shouldn't the relationship to sound quality of this harmonic spectrum explained here? (3) A BJT produces a monotonically decaying harmonic distortion spectrum too (think Taylor series with that factorial in the denominator), so what exactly is the point of this statement? Alfred Centauri 22:49, 8 April 2007 (UTC)

  • 1 Should it say smoothly decreasing amplitudde of harminics instead?
  • 2 Yes. what is it?
  • 3 Who knows?

--SlipperyHippo 22:57, 8 April 2007 (UTC)

The point of this statement is that the "sound" (the subject of this article) commonly (if not nescessarily rigourously) associated with "valves" is in part due to the monotonically decaying dist spectrum that results from class A triode operation (which was the dominant mode of operation during period that most amps used valves, ie prior to the 1950's). The article goes out of its way to make clear that the "valve sound" is in many respects due to the circuits etc used rather than the use of valves per se, using a FET or BJT in class A will indeed also produce this result, but that doesnt invalidate the statement that triodes do it to.
I wrote much text tryingto say something about the relationship between sound quality and distortion, however a user known as light current (now banned) was determined to vandalise teh page endlessly in defense of the transistor and just kept cutting it away (amoung many other irritations). This is actually almost imposisble to address rigourously (at leats to get any agreements) but on a high level, most people who know about this would agree that
  • large low order dist products tend to mask higher order ones, especially if of lower amplitude. hence the significance of monotonicity in the dist spectrum - it means that while the higher order products may be there, the ear/brain more or less doesnt hear them
  • even order products just sound like a "chord" - musical and adding richness and warmth. Odd order sould less musical
  • aharmonic products tend to sound very unmusical / "harsh" etc
by extension combining these, simple monotonic and even order dominated harmonic distortion spectra (as produced by class A triodes etc) sound good. complex distortion spectra (due to complex circuits having large amounts of feedback etc) may have far lower measured peak distortion levels but tend to sound much worse
tubenutdave 23:00, 9 May 2007 (UTC)

It's too bad that Light Current (AKA SlipperyHippo) was banned. I've been plenty frustrated with him at times in the past but I will confess that even so, in every case I came away from those debates with a deeper understanding of the subject matter simply from the excercise of rebutting his occasional odd ideas. Perhaps he crossed the line here - I dunno. The bottom line is this - there is much in this article that is pseudo-science (AKA 'voodoo', 'witch-craft' whatever you want to call it) and I intend to be properly skeptical about such statements in this article and so I will require rigorous justification from the author(s) of these statements. I can assure you that I will provide rigorous justification for mine. Alfred Centauri 02:07, 12 May 2007 (UTC)

Disputed

(this will be the home for disputed sections of this article) Alfred Centauri 23:05, 8 April 2007 (UTC)

(1)

Valves have much less gain and thus permit much lower feedback margins, but have inherently better open loop (i.e., prior to feedback) linearity.

I think the original author was comparing the transconductance of valves with transistors. Are they not generally lower?--SlipperyHippo 23:13, 8 April 2007 (UTC)

That's not the part of the statement that is in question. First, why would the lower gain of a valve permit lower feedback margins? Does the higher gain of a BJT prevent lower feedback margins? Second, why would the lower gain of a valve result in inherently better open loop linearity? Alfred Centauri 23:38, 8 April 2007 (UTC)

I think he meant to say that you dont need as much NFB with valves to get the desired stage gain. However, its not the lower gain of the valve that gives the better linearity, its the valve transfer characteristics themselves isn't it?

--SlipperyHippo 00:02, 9 April 2007 (UTC)

i was the author of that so let me try to clarify what I meant ;-) ..
1)Please NB I at no point wrote that the open loop linearity of tubes was "inherently" due to them having less gain, dont beat me up for things I didnt write !.... It simply turns out to be the case that (especially triodes) are extremely linear, and (especially triodes) tend to have lower transconductance. The reasons tubes are linear is due to the physics of how they function and the geometry of thie structures.
2) By "permits" lower feedback margins I meant simply that (by definition) the feedback margin you have is the relationship between the open and closed loop gain. If you have an active circuit that only has low (open loop) gain to begin with, you have few options - either you have little or no feedback, or have to settle for litle or no closed loop gain. You cant spend more money than you have
tubenutdave 22:47, 9 May 2007 (UTC)

NFB is not needed to set stage gain though it can be used to do so. I agree that the lower gain of the valve isn't relevant to the linearity. Have you written down the Taylor series expansion for

and compared it the Taylor series expansion for

With no NFB and no unbypassed cathode resistor, Stage gain (transconductance) of a triode:
(A) = μRL/(RL + Rp) where μ = amplification factor. So if Rp<< RL, A-> μ.--SlipperyHippo 17:06, 9 April 2007 (UTC)
This appears to be the approximate case, assuming that mu doesnt depend on anything else. However, the correct equation to use appears to be:
as this considers all the variables.--SlipperyHippo 17:11, 9 April 2007 (UTC)

Yes, but look at the Taylor series. Do the higher order terms of the triode function terminate at some finite order or are there an infinite number of terms? How fast do these higher order terms go toward zero? What does this say about the 'inherent linearity' of the triode versus the BJT? Alfred Centauri 00:16, 10 April 2007 (UTC)

OK I will look at this when I have chance.
Its years since I did this stuff at school. But what I can make out is that:
the exponential expansion (about zero)is the series:
1+ x + x^2/2! + x^3/3! + x^4/4!.......These harmonics therfore die away quite quickly (as n!)but there are an infinite number of them
I'm confused with the expansion of (x)^ 3/2.. I tried following the Taylor series formula, but I dont get increasing powers of x in the series becuase the differentials are also powers of x and when you add the indices, all the terms come out to (x)^3.2. Can something to the power of 3/2 be expressed as a Taylor series? Should I be making the expansion about zero?-- 14:22, 10 April 2007 (UTC)gott logged out--SlipperyHippo 14:23, 10 April 2007 (UTC)
Just looked at a copy of Terman (electronic and radio engineering). THe whole valve distortion thing seems very complex. I dont know if there is a page on wikipedia explaining it more simply.--SlipperyHippo 15:05, 10 April 2007 (UTC)

Make the expansion about the quiescent value of v_GK in terms of the differential v_gk, i.e., v_GK = V_GK + v_gk. Here's a stab at it:

So, unless I've made a boo-boo, the Taylor series for i_P is an infinite series. Now, compare the coefficients of (v_gk)^n to the coefficients of (v_be)^n for the BJT series. Alfred Centauri 16:34, 10 April 2007 (UTC)

The first few terms are:

Alfred Centauri 20:35, 10 April 2007 (UTC)

Excuse my ignorance, but is this correct?
or should it be
?
--SlipperyHippo 00:17, 11 April 2007 (UTC)

Good catch! Please excuse my sloppiness. Alfred Centauri 01:52, 11 April 2007 (UTC)

Excused. But as penance, you must answer your own (rather difficult for me, because my math is exteremly rusty) question as to how the harmonics of a valve stage compsre with those of a transistor stage (no feedback whatsoever-- so no emitter followers etc)--SlipperyHippo 12:53, 11 April 2007 (UTC)
I suppose one would need to do the expansion of exp(x) around a convenient bias voltage again?--SlipperyHippo 12:55, 11 April 2007 (UTC)

You suppose correctly and the BJT expansion is much easier than the triode expansion:

Alfred Centauri 19:43, 11 April 2007 (UTC)

OK Well at least I got that right (see above) but I did the expansion about zero. Its interesting to note that we arrive at the same answers (having expanded about different points)
Now the answer as to distortion is plain to see in the expansions! The valve harmonics die away more quickly than the BJT ones. BUT.. one would never use a BJT in this mode for linear amplification. ie valves are to voltage drive as BJTs are to current drive (or equivalent). Agree? Also its interesting to note that the 2nd harmonic magnitude for both are nearly equal whilst the valve wins out at higher harmonics.--SlipperyHippo 21:06, 11 April 2007 (UTC)

It turns out not to be quite that simple, SH. Here's a hint. Let x = A cos(wt) for both series and crank through the math. Each term in the series produces a series of harmonics, not just one. So, to really see how the harmonic distortion components 'add up', you have to crunch some numbers. Here's a hint: each odd power of x produces a component at the fundamental. In the case of the triode, odd powers greater than 1 actually subtract from the fundamental whilst they add for the BJT. So, there is some value of A where the amplitude of the fundamental is equal in both series. Then the harmonic components can be compared. Alfred Centauri 23:02, 11 April 2007 (UTC)

Yes of course you are correct. IM distortion rears its ugly head! But 'IM' not feeling too mathematical tonite. I may attempt crunching tomorrow.--SlipperyHippo 23:13, 11 April 2007 (UTC)
Well this makes sense because valves used to be used as mixers did they not? [8]--SlipperyHippo 23:21, 11 April 2007 (UTC)

One other interesting thing to note: for the BJT, the distortion spectrum depends only on x = (v_be / V_T) whilst the triode distortion spectrum depends on the quiescent values of V_GK and V_PK. Looking at the equation for x in the triode series, for a given v_gk amplitude, what do we need to do with V_GK and V_PK to reduce distortion?

Oh, and one other thing I haven't included (yet). These series assume that the PK and CE voltages are held constant. With a plate/collector resistor in series, the BJT gain is merely reduced by the (linear) Early effect. For the triode however, we get addition non-linearity:

Ouch! Alfred Centauri 23:26, 11 April 2007 (UTC)

BTW, we must be careful not to be accused of indulging in Original research in this talk. The question is: where does common sense/mathematical obviousness cease and WP:OR begin? --SlipperyHippo 22:42, 11 April 2007 (UTC)

Are the OR police in town? Alfred Centauri 23:02, 11 April 2007 (UTC)

They are always lurking (just around the corner)--SlipperyHippo 23:13, 11 April 2007 (UTC)
I will stay out of the discussion on expansions ;-) .. and simply suggest that if you wish to consider the proposition that tubes are / are not more linear than eg FETs, you cant simply look at the expansions (which are only approximations anyway) as algebra - you need to drop in measured values. Alternatively make eg a 300B and a single FET gain stage and then connect it to an audio precision 1 or whatever and measure what the resulting distortion is. ;-) As I have understood it is generally accepted that triodes are extremely linear. And conversely I have not seen ANY transistor / fet circuits that do not have any applied feedback at all (please dont reopen the argument about when is local degeneration feedbac here, thats been battered to death elsewhere and isnt in dispute) tubenutdave 22:47, 9 May 2007 (UTC)

You most certainly can look at the expansions even if these equations are approximations. The fact is that these equations are derived from the physical principles upon which the devices function. Yes, certain simplifying assumptions have been made to derive these equations. But so what? There is no analytic solution to the three-body problem in gravitation yet we've gotten really good at navigating around our neighborhood in the solar system despite this fact. That fact is that the non-ideal aspects of these devices are modeled with pertubations to the fundamental equations above.

Triodes are not "extremely linear" by any definition. Look at that 2nd order term in the Taylor series expansion. Maybe the problem here is definition of terms. What exactly do you believe extremely linear means? To me, extremely linear implies that the higher order terms can be neglected (that is, they are orders of magnitude lower than the linear term). This is certainly not the case for a triode. Triodes generate generous amounts of 2nd order harmonic distortion. What is linear about that?

That fact that you haven't seen any (audio related) transistor circuits without feedback at all doesn't mean that they don't exist. Here's one of many: [9] Alfred Centauri 13:46, 10 May 2007 (UTC)

(2)

Audio valves typically have only modest gain. This makes it possible to design very simple valve circuits that rely on this inherent open-loop linearity and have little, or no negative feedback, and thus have very simple distortion spectra.

Less NFB = nicer sounding distortion?--SlipperyHippo 00:09, 9 April 2007 (UTC)

Possibly, but that's not the point. Why would 'modest gain' make it possible to design simple valve circuits? Why would 'modest gain' imply inherent linearity?

I think its obvious that the author does not like NFB, and thinks it degrades the sound! most of the other statements  ::seem to follow from this belief.--SlipperyHippo 17:14, 9 April 2007 (UTC)

I agree with your observation, but once again, that's not really the point here. These statements tell me that not only does the author not 'like' NFB, he or she doesn't understand NFB or how it is used. Look, the author says that 'because valves have modest gain' ... 'simple circuits are possible'... 'with little or no NFB'. Yet, in reality, simple circuits are possible with little or no NFB regardless of the whether the gain of the device is moderate or high. That fact is, the 'modest gain' of the triode is a limitation. It's simply not correct to say that it makes possible low NFB designs. Rather, it would be more correct to say that the modest gain of triode limits the designer to low NFB designs for 'simple' circuits. Alfred Centauri 21:12, 9 April 2007 (UTC)

Tend to agree ATM --SlipperyHippo 01:03, 10 April 2007 (UTC)

(3)

Another problem is that the regulator is effectively placed inside the signal path, and no real world regulator is ever perfect, it (by definition) can only try to correct to an error in the desired output voltage after the error has occurred. Many arguments have raged regarding if regulators are good or bad.

(4)

It is relevant at this point to note that the class A differential stage draws almost constant current regardless of the signal. The rail pull down is thus constant. This stage has a very high "commmon mode rejection ratio" and good regulation becomes unimportant.

This semms to be saying that because Class A diff stages draw a constant current from the supply rail under all signal ::conditions, you dont need to regulate the rail voltage (which BTW would harm the sound because the 'regulator' is in ::the signal path)--SlipperyHippo 17:19, 9 April 2007 (UTC)

What is a class A differential stage? Is the author referring to a class A push-pull stage instead? To the best of my knowledge, all differential amplifier stages are 'class A'. But CMRR is a performance measure of differential amplifier stages. But then again, it is PSRR that determines if good regulation is important, isn't it?. All in all, it seems to me that the author of those sentences is terribly confused. Alfred Centauri 20:56, 9 April 2007 (UTC)

Thinking!--SlipperyHippo 01:03, 10 April 2007 (UTC)
No not quite all! You could have a bridged amplifier where each half is class AB. This could be said to be a differential 'stage'.--SlipperyHippo 23:32, 10 April 2007 (UTC)

A good point! On the other hand, in order for there to be any current through the load at all, one side must 'push' while the other must 'pull' and this must be so over the entire operating range, right? So, I would argue that, regardless of the operation class of the individual sides, the bridged output stage must be considered class A (neither side is 'off' over the operating range). Thoughts?

Class A made up of 2 class (A)B stages. Yes, I suppose--SlipperyHippo 12:41, 11 April 2007 (UTC)
But there again, if you want 360 deg angle of flow (conduction angle), in the end everything has to simulate class A doesnt it?--SlipperyHippo 13:31, 11 April 2007 (UTC)

(5)

The above simple analysis only applies to single tone continuous sine waves. Real world music contains transients and many tines at once. When multiple tones are present, if the amplifiers transfer function is not perfectly flat (which it never is in the real world), intermodulation products will also be generated (these typically being harmonic to the original tones involved.

The author seems to be confusing transfer function with non linearity as far as I can see.
--SlipperyHippo 17:19, 9 April 2007 (UTC)
Something that is non-linear will have a non-linear transfer function. Transfer functions are almost always assumed to be linear so you can do LTI stuff, but they aren't linear in real life, and obviously aren't assumed to be when you're talking about non-linearities. See Talk:Transfer_function#LTI_systems_only.3F from a long time ago. — Omegatron 19:36, 9 April 2007 (UTC)
Depends what you mean by TF--SlipperyHippo 01:03, 10 April 2007 (UTC)

I was under the impression that a transfer function 'lives' in the frequency domain. From my perspective, a perfectly flat transfer function (magnitude) implies perfectly flat frequency response. To the best of my knowledge, the only non-flat linear transfer function is of the form Af corresponding to a differentiator in the time domain. Any other transfer function is a non-linear function of f but that doesn't imply non-linear distortion in the time domain. In fact, a non-linear system doesn't have a transfer function in the same sense as a linear system does. For a linear system, the transfer function is simply the frequency domain representation of the time domain unit impulse response.

Back to the quoted text from the article: I believe the author is referring to the time domain transfer characteristic curve instead of transfer function. Yet, a transfer characteristic that is flat is and non-zero would certainly characterize a non-linear system. For the system to be linear, the transfer characteristic must pass through the origin and must be a line (linear). If the line passing through the origin is flat, then the gain of the system is exactly zero.

Further, intermodulation products are certainly not typically harmonically related to the original tones. Intermodulation components are at frequencies that come from sums and differences of the original tones. Alfred Centauri 00:12, 10 April 2007 (UTC)

agreed--SlipperyHippo 01:03, 10 April 2007 (UTC)

:I admit dont understand electronics to the same depth as you, but what about Af corresponding to an integrator in the time domain? why would that be nonlinear? --SlipperyHippo 01:03, 10 April 2007 (UTC)

Af is a linear function of frequency and corresponds to a differentiator in the time domain which is a linear system. However, a time domain integrator is also a linear system but the frequency domain representation is A/f which is certainly a non-linear function of frequency. Do you see? A linear system doesn't imply a linear transfer function (by transfer function, I mean frequency response). Once again, the transfer function lives in the frequency domain and is the frequency response of linear system. Alfred Centauri 01:17, 10 April 2007 (UTC)

:No I dont quite see yet. Youre saying a diffr is linear and an intgerator is non linear? Yes? --SlipperyHippo 01:24, 10 April 2007 (UTC)

Let me repeat myself: "A linear system doesn't imply a linear transfer function (by transfer function, I mean frequency response)". An integrator is a linear system, right? Look at the transfer function (frequency response) of an integrator. Is it linear? Once again, a linear time domain system does not imply a linear transfer function.

Maybe this will help. Give me an example of a transfer function for a non-linear time domain system. Alfred Centauri 01:43, 10 April 2007 (UTC)

Sorry I was thinking of the bode plot (in dB) of an intergrator. On a linear scale, the TF of an integrator is of course non linear. TF of a differentiator is linear--SlipperyHippo 10:16, 10 April 2007 (UTC)

Audiophile admission?

Is the following an admission by audiophiles that its ok to mess with the recorded signal by adding their own favored brand of distortion by means of a (valve) amplifier of their choice?

However at the upper end our audio systems (so called "high end" or "audiophile" systems) it is recognised that "accuracy" is not something that can be simply described by eg a wide quoted frequency response and a low measured distortion level (THD) - the music also has to sound "natural", and in fact "musical", rather than eg "sterile".

All real world designs distort to some degree, and designers seek a compromise in the nature of the distortion their design produces, that "sounds good", rather than simply measures well. In the high end world audiophiles will typically pay greater regard to how it sounds than to how it measures, to the point of just not caring about measured distortion levels at all.

--SlipperyHippo 15:02, 11 April 2007 (UTC)