ESP Logo
 Elliott Sound Products Valve (Vacuum Tube) Myths 

Valve (Vacuum Tube) Myths

Copyright © 2009 - Rod Elliott (ESP)
Page Created 09 Nov 2009

HomeMain Index ValvesValves Index

We are all used to myths, and they appear in all aspects of life.  Ranging from emails that tell you your house will catch on fire if you open an email from <insert name here> to claims that anyone can fix their own valve guitar amp, they are sometimes merely annoying, but for the latter topic at least, are extremely dangerous.

You can fix your own guitar amp, provided you are an experienced technician who knows what to look for and can diagnose faults without resorting to applying power to see where the smoke comes from.  While a valve change may seem simple enough, it's not, and valve suppliers make matters much, much worse by claiming it can be done - indeed should be done by the owner.

This article covers most of the more popular myths, but I'm sure I will have left something out.  If you fail to find your favourite myth here, please let me know.

Every so often, one comes across claims that simply defy belief.  The following (in context but with specific references removed) caught my eye ...

Without exception, the above claims are - quite bluntly - bullshit!  These comments imply that transistors (including FETs) and opamps can't provide the so-called benefits listed, and that's absolutely false in all respects.  If any of these claims were even a tiny bit true, modern test equipment would be able to detect the difference between a valve and transistor stage.  In fact, the test gear can detect the difference - valves are noisier and have higher distortion than an 'equivalent' (and competent) 'solid-state' circuit.

It is possible to engineer a valve preamp or power amp to (almost) equal a transistor design, but it will be disproportionately expensive, heavy and will liberate (comparatively) vast amounts of heat.  Valves have a very finite life ranging from minutes to many years, depending on how they are (ab)used.  Replacements are expensive and are now often of dubious quality.

It's worth pointing out one fact that you probably won't find elsewhere.  There is one supply voltage in a valve amplifier that's more important than any other.  That's the negative bias supply for fixed bias push-pull amplifiers.  That is the one voltage that must never fail, for any reason.  If the high tension (B+) supply fails, the amp doesn't work, but nothing bad happens.  Likewise for the heater supply (although failures are very uncommon).  The only thing that prevents the output valves from drawing their maximum possible current is the negative bias supply, so it should always be over-engineered.  Unfortunately, this is very rarely the case, and it's commonly designed to be 'good enough.  This is one place where 'good enough' just isn't good enough!

2 - Myths

The myths surrounding valves are many and varied, but they fall into a number of major categories as shown below.  The major myths involve the 'quality' of the distortion produced, and while there most certainly are differences between valves and transistors (or ICs), most of the stuff you will see is drivel - it's not true at all.  The points listed above are examples of the deluded thinking that plagues the audio world.  Let's go through the list ...

2.1 - Distortion

The most common claim about valve amps is that they have 'nice' second (or even) harmonic distortion, whereas transistor amps have 'nasty' third (or odd) harmonic distortion.  This is true in only the most limited number of cases, and is especially noticeable with single-ended amplifiers.  The fact is that it's not the order of the harmonics (odd or even), but how far they extend beyond the original frequency.  High order harmonics (typically from the seventh and above) do sound relatively nasty, and that applies for both odd and even harmonics.

For push-pull amplifiers, clipping is usually symmetrical.  If properly designed for low distortion (hi-fi amps), there will be small amounts of both even and odd harmonics, but they will be limited to perhaps the fourth or fifth, with diminishing amounts of anything higher.  Once the amplifier clips, the distortion produced is predominantly odd-order, and for a valve hi-fi amp with significant feedback, it will probably sound remarkably similar to a transistor amp clipped to the same degree.

For guitar amps, most of the input stages are biased to provide maximum symmetrical swing, which means that the distortion will be a mixture of even and odd harmonics.  At (and beyond) clipping, the distortion is predominantly odd order.  Strangely, very few guitarists actually like asymmetrical clipping, which will always be a combination of even and order.  Severely asymmetrical clipping usually sounds dreadful - it develops a tonality that may be described as 'thin and reedy', or 'farty', depending on the severity and the way the amp is used.  The allegedly smooth sound of even order distortion only (particularly the second harmonic, which is claimed to be especially nice), produces a waveform that still tends to look like a sinewave, but just isn't quite right.  At anything above 10% distortion, it really doesn't matter whether the distortion is odd or even order - for hi-fi this is simply unacceptably high.

To prove these assertions is easy, as the following waveforms will show.  Unfortunately, there are some people to whom there is no proof, because they 'believe'.  If you believe the nonsense that marketing pukes and vested interests dream up, then no proof will work, but if you are willing to see exactly how odd and even harmonics create new waveforms, then you need to look at the waveforms.  Due to space and article size considerations, I've only added in the first harmonic of the fundamental, at 10% distortion.  In the case of a typical valve circuit approaching clipping, the first odd harmonic is the 3rd, and is 20dB lower than the fundamental.  1V at 400Hz + 100mV at 1.2kHz ... the waveform is shown in Figure 1.

fig 1
Figure 1 - Symmetrical Waveform at the Onset of Clipping

For even order distortion, the typical waveform seen from a valve preamp circuit looks somewhat like that shown in Figure 2.  This waveform is a mix of 1V at 400Hz + 100mV at 800Hz.  The second harmonic is displaced by 90°, for the simple reason that this is necessary to recreate the waveform seen.  The valve itself doesn't somehow 'magically' generate a 90° phase shifted sinewave, it distorts the signal, and the harmonic(s) are generated by the act of distortion.

fig 2
Figure 2 - Asymmetrical Waveform From Single Ended Stage

The distortion is actually quite difficult to see if you're not used to looking at sinewaves, so the second half of the graph shown includes the fundamental superimposed in green.  This makes it easy to see that the top of the waveform is slightly squashed, and the bottom is stretched.  At less than about 5% distortion, it is extremely hard to see any difference in the waveform, but naturally a distortion meter or FFT will show that it's there.

Audibility is another matter.  With a sinewave, it's usually easy enough to detect as little as 0.5% distortion reliably.  With music, that level is unlikely to be heard unless the loudspeakers are particularly revealing, and the listener has a keen ear.  Even much higher levels can be difficult to detect by ear, and this again depends on the speakers and the type of music.  Note that these comments are intended as a guide only, and do not include the effects of intermodulation.  IMD results from any non-linearity, and is measured with two frequencies.

A common test is to apply signals of (say) 1kHz and 1.1kHz.  IMD is measured by looking at the amount of 100Hz (and 2.1kHz) signal produced, since IMD produces sum and difference frequencies (provided the distortion is asymmetrical - see note 1), as well as the normal progression of harmonics (and their sum and difference frequencies).  As a result, minimising IMD is essential, or the end result can sound revolting.  To demonstrate this, look at Figure 3, which shows a simple distortion circuit.  When operated with a single 1kHz signal, distortion is about 7% and consists of both odd and even harmonics.  When a second signal of 1,100Hz is provided as shown, IMD can be plotted as shown (the chart is the result of a FFT in the simulator).  The completely new frequencies at 100, 200, 300, 900 and 1200Hz are quite visible, and only intermodulation can create that.  Look at all the other frequencies shown too.  With a single frequency, we expect a diminishing sequence of harmonics - even and odd.

fig 3
Figure 3 - Asymmetrical Distortion and Intermodulation

With IMD, we see new sets of frequencies centred around 2.1kHz, 3.15kHz, 4.2kHz, and extending well past the end of the graph.  Only a few of these are actually harmonically (and musically) related to the original two frequencies, so the resultant distortion is often (highly) discordant.  Note that the sum and difference frequencies are not generated with perfectly symmetrical distortion (strange but true).

1     Symmetrical distortion produces intermodulation products, but not the sum and difference frequencies.  This fact seems not to be well known or understood.

If you were to build a valve preamp operated at a low voltage so that it will generate significant distortion at low levels (as I have done), it's very easy to dispel any myth that second harmonic distortion sounds 'nice'.  It doesn't!  My circuit was biased so that it would generate about 3% distortion with an input level of 300mV (this is the level I get from my FM tuner in my workshop system).  The distortion at that level is almost exclusively the second harmonic, and a fairly clean sinewave at double the input frequency was observed as the distortion residual at the output of my distortion meter.

According to those who think that such distortion levels give 'body' or 'something' to the sound, this arrangement should have sounded 'pleasant' with music - much as one might expect from a SET (single ended triode) amplifier operating at normal levels into a typical loudspeaker.  Well, it certainly gave the sound 'something', and that is best described as crap - it sounded bloody awful.  There was clearly audible intermodulation distortion that cluttered the sound, and the overall effect was dreadful.  No magic sound here - simply distortion that was audible even on speech.  Reducing the level reduced the distortion too, but it never sounded as good as the direct signal without the added distortion.

I suppose one could get used to the sound - after all, in the (early) 1930s that's what people had, they got great enjoyment from it and there was nothing else available at the time.  Even by 1936, many manufacturers had changed to push-pull output stages with output power in the range of 10-20W for their top-of-the-range wireless receivers.  A lot has changed since then, and we are now able to listen to music without audible distortion.  Given that it's possible to make a direct comparison between a distorted and undistorted signal with ease (any ABX test would almost certainly give a 100% reliable differential), I simply cannot understand why anyone would ruin the sound from vinyl, CD, SACD, etc. by feeding it through an amplifier that is guaranteed to produce significant harmonic and intermodulation distortion.  It doesn't sound nice, and designers have been working for decades to produce amplifiers that have distortion levels that are below the threshold of audibility.

Engineers didn't spend all that time and effort to make amplifiers sound worse, that much is certain.  There are some simply outstanding amplifiers available today, with specifications (and sound) that was unheard of in the (mercifully brief) heyday of the SET amplifier.  In reality, the era of SET amplifiers was rather short, because it was quickly found that a push-pull amp was vastly more efficient, had significantly lower distortion and more output power, and could use a smaller transformer for better performance at both high and low frequencies.  No sensible designer of that period would even consider a single-ended amplifier of any kind, except for use in mantel radios and other comparatively cheap consumer products.  Such amplifiers were common in mantel radios, 'radiograms' (combined radio and record player) and TV sets up to the 1960s, after which they thankfully faded into the obscurity they so richly deserved.

It's probably worth noting that some of the most expensive amplifiers of the valve era were designed for very low distortion (such as the Leak TL/12, with claimed distortion of 0.1%), and their initial (or primary) purpose was to drive the cutting head in disc cutting lathes.  No-one would ever think it was a good idea to use an amplifier with high distortion to cut vinyl masters, and low distortion amps would naturally be preferred.  Obtaining distortion levels below 0.1% means that good design and plenty of feedback is needed, because no amp (valve or transistor) can achieve the required performance without generous negative feedback.

2.2 - Linearity

Many is the claim that "valves are linear, but transistors are not".  This is bollocks!  Valves are not linear, and the simple proof of this is a glance at any plate voltage vs. current characteristic chart.  If the valve were in fact linear, there would be a series of straight lines on the chart, not the curved lines we always see.

Part of the reason for this myth is easy to explain.  When valves and transistors competed in parallel for the mass market, we had very basic circuitry in both types of amp/ preamp.  The difference was that the valve equipment operated from 300-400 volt supplies, while the transistorised equivalents used a supply of 20-30 volts.  A signal of 1V RMS (2.8V peak-to-peak) occupied only 0.93% of the total available 300V supply, while the same signal in a transistor used 10 times this (9.3%) of a 30V supply.  Higher signal voltages were obviously worse in this respect.  Added to this was many years of experience with valves, and very little experience with transistors, so most transistor circuits were probably sub-optimal.  The early transistors were also a far cry from those we can get today, with lower gain, higher noise, etc.

Added to this were some fundamental (glaring) errors made when people started using transistor circuits.  One that particularly amused me was the first (non-adjustable) gain stage in some early transistorised microphone preamps.  It was traditional that a high gain valve would be the front end, and these typically had a voltage gain of about 50.  With high impedance microphones being the normal ones used at the time, almost anyone could get a 1V or more quite easily from the mic.  These mics used a normal low impedance cartridge with a step up transformer.

However, even up to 1V can be obtained from a low impedance mic right against a loud singer's mouth quite readily - I've tested this, and managed it with ease.  The valve would have some difficulty (trying to get 50V RMS, or about 140V p-p), but distortion was not horrendous.  When transistor circuits started to be used, the gain (also non-adjustable) of the input preamp was ... 48 (nominal - based on a 47k and 1k resistor).  Since this operated from a 30V supply, anything over 200mV input caused gross distortion.  Silly decisions like that were not uncommon in the early days, and each such mistake provided ammunition for the anti-transistor terrorists.

Despite that, an early (basic) single transistor stage with emitter degeneration (local feedback) could still more than match most valves for linearity, and as transistor circuitry improved it became no-contest.  High quality opamps today can achieve distortion levels (a direct indicator of linearity) that are almost unmeasurable.  No valve circuit can even approach the distortion - both simple harmonic and intermodulation - that can be achieved easily with a $2 opamp.

If the collector current in a transistor is plotted for varying collector voltage at a fixed base current, the 'curve' is almost a straight line - far more linear than any valve.  Several other tests give a very linear result too - but ... gain varies with emitter current, and that relationship is decidedly non-linear, especially at low current.  A valve is no different in this respect, and the reasons even have a tenuous relationship between the two devices.

fig 4
Figure 4 - Linearity Test Circuits ... Transistor vs. Valve

Now, my claim that a transistor in its simplest configuration will beat a valve is bound to have a few people shaking their heads sadly.  I can almost hear the laments as I write ... "Poor silly bugger, he's really lost it this time." Sorry to disappoint, but I tested this with real parts, and was surprised at the results.  Very surprised, actually.  The transistor circuit is much, much better than the valve, despite the huge supply voltage difference.

I added the cathode follower to the valve circuit because without it, I couldn't drive my distortion meters properly because of the high output impedance, and I do admit that has a small effect.  However, I re-ran the test with two different distortion meters ... one (the 'preferred' unit) has an input impedance of 100k, and with that connected directly to the plate circuit, output voltage was reduced and distortion increased to 0.8%.  The other meter has an input impedance of around 1M, and that showed almost identical distortion from the plate of the amplifier and the output of the cathode follower.

The valve distortion at a very modest output voltage (1.21V RMS) was a reasonably respectable 0.4%.  Not wonderful, but pretty much as I expected.  I know that this can be improved, but this is a simple test of linearity, and the valve stage is set up for symmetrical clipping for maximum dynamic range.  The distortion climbs steadily as output voltage is increased though.

So what of the transistor, also biased for symmetrical clipping with the voltages shown?  With exactly the same gain (hence the odd value emitter resistor) and voltage, I measured 0.13%, which like the valve, was predominantly second harmonic.  Biasing the transistor for a collector voltage of 4V reduced the distortion to less than 0.06%, and although interesting, it's not useful because there is no dynamic range before distortion goes through the roof.

For both circuits, some of the meter's output residual (distortion + noise) signal was noise.  The valve stage has a reasonable earth plane, but the transistor circuit was literally just hanging in mid air, supported by test leads running all over my workbench.  Looking at the residual waveform on my oscilloscope, most of the signal from the transistor stage was noise, and it was necessary to use averaging to see the 2nd harmonic signal.  This indicates that the distortion is actually lower than measured on the meter!  Some of the noise was picked from a fluorescent lamp that's no more than 300mm from where I was working, but there was little evidence of this noise from the shielded valve circuit.  This has good supply decoupling, rather than 1m leads to a couple of power supplies as used for the transistor test.  Overall, a very unfair test on the transistor, yet it still won easily.

Naturally, these are tests that anyone can do.  You do need a low distortion oscillator and a distortion meter (or use a PC oscilloscope with FFT capabilities), and you can expect similar results.  I used a BC546 because I have lots of them here and I used the first one out of the bag.  No selection or tweaking was done, apart from adjusting the emitter resistor with a decade box until the gain was the same as the valve stage.  I also ran a simulation, using exactly the same values as the real transistor circuit.  The simulator claimed a distortion of 0.1%, so has obviously under-estimated reality (as I expected).

The test described here isn't designed to do anything more than show that both valves and transistors are non-linear, and in particular to dispel the myth that valves are linear.  It's not meant to prove that transistors are 'better' than valves, but more to show that both will show non-linearities under similar conditions of operation.  Both circuits have local feedback, but the transistor has more simply because it has higher gain.

2.3 - Sinewaves are Simple

This myth applies to both valves, transistors and ICs, but it's still worth including here because it's a common claim - especially by subjectivists who cheerfully refute any measurement data.  Many people keep complaining that sinewave testing is 'too simple', and that it's somehow easy for an amp to reproduce because of this inherent simplicity.  A sinewave is simple, only in that it is a mathematically pure signal, and contains exactly (and only) one frequency - the fundamental.  Sinewaves have been used for testing for decades, and any amplifier that can reproduce a sinewave perfectly therefore has zero distortion.  If harmonic distortion is zero, it follows that there is no non-linearity whatsoever, so IMD will also be zero.

No amplifier ever made can reproduce a sinewave perfectly, regardless of its topology or technology.  Some do get very close, but once distortions are below 0.01% it will usually be impossible to pick one from another, provided frequency response and gain are also identical.  As I've pointed out in several articles, amplifiers have no idea of the signal waveshapes they are reproducing.  All that matters is that at any instant in time, there will be one specific voltage that must be amplified, and if that is amplified by the same amount regardless of the instantaneous input voltage, the amp has no distortion.  Should the signal cause the amp to attempt to provide an output signal that's greater than the supply voltage, the amp will clip the output - this is clipping distortion.

Signals that change too fast for the amplifier to keep up will create distortion.  This effect was dubbed TID (Transient Intermodulation Distortion), and it can be induced in any amplifier ever made, including those that don't have any TID according to their designers.  All you need to do is apply a signal that changes so fast that the amp is incapable of the required rate-of-change at the output (slew rate limiting).  My test oscillator can produce a squarewave that's faster than any amp I've tested or read about can possibly handle.  Big deal - so can hundreds of other squarewave generators.

What was missed (or the authors chose to ignore) in the early TID papers and claims is that while oscillators can generate these signals, music can't!  There is simply no traditional musical instrument that can generate transients anywhere near as fast as an electronic oscillator.  Even synthesisers won't do it, because they have filters (and opamps) within the circuitry that limit their ability to create extremely fast transients.  This is necessary, not only to make them compatible with amplifiers, but because such fast transients can't be heard (or reproduced) anyway.

CD music is brick-wall filtered at 21kHz, vinyl is utterly incapable of transients that will stress amplifiers (such a transient would likely tear the stylus off its cantilever, and can't be cut into the master).  Higher sampling rates don't change anything either.  Microphones can't react to an instantaneous step change in pressure, and even if something could create it, the air carrying the signal won't support it either.  There are scientific devices that can do it - they use a disk designed to rupture past a certain pressure to generate massive pressure gradient waveforms.  I don't know of any musical piece that uses them, and no, the 1812 Overture's cannons don't count.  The cannon risetime is actually quite leisurely, and won't stress anything apart from loudspeakers that can't handle the peak power at low frequencies.

In a sense, sinewaves are simple, and it is their very simplicity that makes people think that they must therefore be an 'easy' signal to reproduce.  I can easily hear 0.5% distortion on a sinewave with my workshop system (horn loaded), and a more revealing system may do better.  Trying to hear the same level of distortion on music is much harder - especially if it's only on transients.  In that case, most people wouldn't hear it at all in isolation, and even in a double-blind test may have difficulty.

I've also seen it claimed that the speed of electrons in a vacuum is much greater than through silicon, so valve amps are somehow 'faster' or perhaps more 'immediate'.  What a load of unmitigated horse feathers!  If you see such a claim, it should be immediately obvious that the author is either mad (i.e. clinically insane, typically delusional), lying, or also believes that there are fairies at the bottom of his garden (or between his ears) that make his spinach¹ taste better than anyone else's.

1 - If you don't like spinach, please insert vegetable of choice ... as a replacement for the word spinach, not there  

2.4 - Matched Valves

You can buy matched pairs of output valves.  That's what the advertisements say, and they wouldn't lie to you, would they?  Perhaps, or perhaps not, but they just might lie by omission - the omission being that few will tell you the matching conditions, explain the test processes, or disclose what is matched and what is not.  To be properly matched, the valves have to be tested with the plate and screen voltages you have in your amplifier, and operated in a similar manner to that in the amplifier.

While never stated, there is an implication that since matching involves testing, matched valves are somehow better than 'ordinary' unmatched valves.  In reality, it makes no difference if crappy valves are matched or not - they're still crappy valves.  No method of matching, regardless of how rigorous the tests might be, will make a badly made valve work any better or last any longer than it does with no testing.  The occasional completely dead valve might get thrown away, but that's cold comfort.  If there is a significant number of failures in a batch from a manufacturer, then the whole lot should be binned, not just those that fail immediately.  The ones that pass through the matching process are not better built than those that failed, they simply haven't failed ... yet.

Any form of matching test that involves a traditional valve tester is useless, because at best, these testers give nothing more than an indication that the valve is functional.  Scales that simply have red and green sections marked 'Replace' and 'Good' are worse than useless.  Does your supplier match valves using one of these testers?  You'll probably never know, because they don't (or won't) tell you.  Unless there is full disclosure of the voltages used for plate, screen and control grid, and that the valve is matched for a specified idle current and gain, then the matching is as good as arbitrary.  There are so many differences between valves that obtaining a genuinely matched pair, where they track each other over the full operating range, is extremely difficult.

Ideally, you need to be able to match the valves yourself, in your amplifier, and under the conditions as those where they will actually be operated.  Even after this is done, there can be changes in the early part of the valve's life that can cause a significant variation between valves that were originally carefully matched.  As with 'professionally' matched valves, just because you go through the process doesn't mean that the selected valves will work properly over their lifetime or last as long as they should.

Some valves with which I had personal experience were 'premium' KT88s.  After the first 100 hours or so of use, every time the amplifiers were turned on the idle current changed.  After adjustment (the amplifiers had LEDs and accessible trimpots to set the bias) all was well until the amps were turned off and back on.  Sure enough, the bias current was different again.  Initially, I was only able to do a thought experiment on this problem because the system was still in operation and no replacement valves were available at the time (20 of them in all).  I came to the conclusion that the heater was probably loose in the cathode tube, because nothing else explained the fault satisfactorily.  When later I broke one apart, I tested my theory, and the heater wire literally fell out of the cathode.  As a 'sanity check', I dismantled an old (dead) GE version for comparison.  The GE valve's heater had to be coaxed out of the cathode, and was difficult to reinstall because it was so tight.

No matter what anyone may claim, valves such as this cannot be matched.  Every time power is applied, the valve will have different characteristics because different parts of the cathode are slightly hotter than the rest.  In reality, these valves were fit for nothing - certainly they were of no use in an amplifier.  They were faulty from manufacture, but seemed to be alright when first used.  It took some time before the problem really showed up, so any matching process could never have picked up that anything was amiss.  I doubt that this was an isolated incident either.  Quite obviously, we didn't get the world's total supply of these faulty valves, so others would have had the same problem.  Sloppy workmanship, poor equipment setup, material substitutions and any number of other manufacturing faults can play havoc with the performance of any valve, but unless it completely fails your chances of a warranty replacement are very slim indeed.

Treat all claims of matched valves with suspicion.  Before handing over any money, find out the exact conditions used for matching, and if these conditions are significantly different from your amplifier, then don't bother.  Anyone who cannot (or will not) tell you exactly how the matching tests are done should be avoided, because without knowing the details of the process used, the matching is likely to be no more useful than deciding that they are matched because they look the same.

It's less than inspiring to read through a detailed article about how one vendor matches valves, only to discover through the text that there are some fairly obvious errors that show that the writer doesn't understand as much about valves as he thinks he does.  This being the case, can you trust the company involved to know the proper way to conduct the tests?  Perhaps they do, but to include (IMO) serious errors in an article intended to demonstrate that their matching process is better than (unnamed) 'others' doesn't engender much trust from me.  I do understand that they are making the article readable to the 'average' enthusiast, but this being the case it would have been better to leave out the wrong stuff altogether.

One thing is for certain though.  There is absolutely no point whatsoever buying matched valves if you don't know which parameters are matched, or at what voltages.  To make things harder, there is considerable variance of output stage topologies, and some are far more tolerant of a mismatch than others.  Valves that are matched at idle - i.e. for a specific bias current measured under (usually) unknown conditions - may have widely differing mutual conductance (gm) and/or plate resistance (rP) at electrode voltages other than those used for the test.  While it has been suggested that µ (Mu, or amplification factor) should be matched, this generally has to be calculated from gm and rP because it's difficult to measure directly.  It can be done, but for the most part is not actually very useful at all.

Many resellers claim to have 'computerised' valve matching machines.  What they don't say is what the computer does - for all we know the programme simply activates a few relays to connect the appropriate pins to their respective supplies, but nothing else.  It might not even do that - there are plenty of 'computerised' systems that use nothing more than an 8-pin microcontroller to perform very basic logic functions that might be irksome to do by other means.  All of the suppliers who match valves do so for one reason only - to try to convince you to purchase their product.  They generally don't do it because they are really nice people and want you to like them as human beings, so you should treat everything they tell you with suspicion.

2.5 - Valve Sound

Often you will hear claims that a certain brand (or type) of valve is 'sweeter' than another, or has greater 'bass authority'.  These are weasel words that have no defined meaning whatsoever, so the claims are equally undefined and can most likely be attributed to imagination (or vested interest if you like conspiracy theories).  There is no doubt that different valves can sound different, but unless the writer is willing to take measurements that explain the difference, treat all claims as apocryphal.

Unfortunately, there are masses (or so it seems) of people to whom a measurement is worse than an anathema, it amounts to sacrilege or even heresy.  Audio measurements were developed because subjective testing is unreliable, and can be influenced by a great many things that have absolutely nothing to do with the device under test.  To be really useful, measurements have to be conducted properly, and the full details of how the measurement was made explained.  There are a few measurements that don't need much explanation, because the methods are considered well known to the extent that anyone 'skilled in the art' will immediately understand how it was done.

If one pair of valves sounds different from another, then there's a reason, and that reason can be measured.  Despite claims that there are things in audio that simply cannot be measured, these are just more weasel words - usually used by purveyors of magic components and/or snake oil.  Measurement techniques are well advanced, and engineers and scientists know exactly what they can get away with and not have people bitching and moaning - MP3 is a perfect example.  It's rubbish and sounds dreadful compared to CD, but thousands upon thousands of people are perfectly happy to listen to it.

Be that as it may, there is nothing about an amplifier that makes an audible difference (as determined by a proper double-blind test) that cannot be measured, and test instruments are more than capable of resolving differences that are completely inaudible.  Therefore, any claim that one valve sounds significantly better than another (for whatever reason) is meaningless without a series of measurements that show what the differences are.  There is no doubt that there are real differences, and these will show up as lower (or higher) output impedance, distortion (harmonic and intermodulation), etc.

When people refer to 'valve sound' vs. 'transistor sound' in amps, then there are some significant differences, although not all are necessarily audible during normal listening.  Some of these have already been covered in another article and will not be regurgitated here, other than to point out that most valve amps have a relatively high output impedance (low damping factor), so allow the speaker impedance to influence the frequency response.  When demonstrated, higher than normal output impedance is almost always described as sounding 'better', but the listener is often unaware that it is no longer accurate.  This is of no consequence if you really do like the sound of an 8 ohm speaker driven from an amplifier with a 2 ohm (or more) output impedance, and exactly the same thing can be done with transistor amps - I've been doing it for decades.

There are some other differences too, some subtle, others not.  Distortion is an obvious one - no valve amp ever made can equal the distortion figures (or distortion inaudibility) of even typical quality transistor amps.  Crossover distortion is very audible with transistor amps, and less so with valves, but there is no reason or excuse for any amp to have audible crossover distortion, regardless of whether it uses valves, transistors or a combination of the two.

Provided we restrict ourselves to hi-fi amplifiers, valves really offer few benefits and many drawbacks.  Duplicating the output impedance and frequency response are easy with transistor amps, but this is rarely done for the simple reason that the wider bandwidth and lower output impedance of transistor amps is now expected, and the low impedance is used as the typical source for loudspeaker designs.  Changing the source impedance for a loudspeaker system changes the design, so low source impedance is assumed.  Always.

Many factors are harder with valve guitar amps.  Much of the 'sound' is the direct result of either marginal (or poor) design of the output transformer and valve output stage, and/or the omission of any means to prevent high voltage spikes in heavy overdrive.  While it is theoretically possible to regenerate (at least to some extent) the results of these (IMO) common valve amp mistakes in a transistor amp, no-one has ever done so.  It's probably safe to assume that it's not been done because the sound can be bloody awful in extreme cases.

There can be no doubt that different valves can have different overload characteristics, and these differences may be audible in some cases.  Very few opinions on the Net have backup by way of blind testing, so they are just that - opinions.  Unless there is validation by way of double-blind comparisons and/or detailed measurements, pay no heed to the claims that EL34s and 6L6 valves are 'chalk and cheese' (for example), or that one brand of output valve sounds 'superior' to another.  If it does, then the difference can be measured.

2.6 - Blue Glow

Rumours persist that any blue glow inside a valve is bad, and indicates a fault.  This is not necessarily the case, and there is some very good information on the Net that describes the different types of glow one might see and what each means.  In particular, I strongly recommend a web search, as unfortunately the site that I had linked here has disappeared.  I think the same page is now at BLUE GLOW IN ELECTRON TUBES.  To search for more information, just search for the same title.

In short, the blue glow depends on many factors, and for power valves it is common.  Much of the cause is fluorescence of the glass when bombarded by high speed electrons, and the glow is typically a very deep blue, and is seen on the inside of the class in places where electrons manage to escape capture by the anode.  Many valves have holes or cutouts in the plate to facilitate grid alignment during manufacture, and these make ideal escape routes for some of the cathode current.  There are other reasons for a valve to glow, and only one type of glow indicates a fault.

Some valves rely on low pressure gas (neon, krypton, etc.) and were used as voltage regulators - this was well before zener diodes were invented.  These are meant to glow fairly brightly, and the colour depends on the gas mixture used.  Far less common, except for high power applications, were mercury arc rectifiers.  These have a characteristic blue hue, and also emit significant ultraviolet light.  While vintage enthusiasts may come across them occasionally, most will never see one.  The only one I've seen was a very large 3-phase version (I seem to recall it was about a 400mm diameter spherical shape), and was probably used to provide DC power for electric trains.

Should you see an output valve with a faint light blue or purple glow inside the plate, then it is probably gassy, and is approaching failure.  The glow is caused by electrons striking gas (air) molecules, causing them to fluoresce.  All valves have a tiny residual amount of gas, but there should never be enough to create the characteristic glow of a gassy valve.

A good indicator that a valve is gassy (rather than just causing fluorescence) is to look at the vaporised 'getter' material - the (usually) silver coating on the inside of the glass.  In most cases, the first indication that there is gas involved is that the edges of the silver getter coating turn brown.  A valve with no vacuum at all (or at least lots of gas) may have a getter that's all brown or even light grey or white (the latter two mean that there's oxygen in the valve - a sure sign that the glass is cracked somewhere).  As a general rule, don't use any valve where the getter deposit is not as it should be.  Its purpose is to absorb stray gas molecules that are liberated from the metal and glass from which the valve is made, and if it looks a different colour from that of another valve of the same type, then assume the valve is faulty.

You Can Replace Valves Yourself

Note that this refers primarily to output valves, and it is usually quite ok to change preamp or other low-power valves without much risk.  Otherwise ...

You can replace valves yourself.  Absolutely.  No question at all.  They are in sockets, and anyone can remove and replace them at will.  Doing so will always give you a result, but it may not be the one you hoped for.  There are so many reasons not to do so that listing them all would be tedious.  Instead, I'll note the things that must be part of a valve change, as well as those things that should be included.  In some cases, the tests will require equipment that the valve amp owner simply doesn't have or even have access to.

Apart from equipment, there is also a level of knowledge (and experience) that is necessary to ensure a satisfactory valve replacement.  Without these two, both amplifier and owner are at risk, and that risk is especially great for the inexperienced owner.  Delving around inside an amp that runs from a 500V (or more) supply is not just not recommended, it's positively dangerous.  One tiny slip of a finger, and your loved ones could be lining up to place flowers on your coffin - and no, I'm not exaggerating - not even a little bit!

Assuming that you survive the ordeal, how will you know that the end result is better, worse or the same as before?  If all you have to go on is your ears, unless the original fault was gross (one output valve not working for example), you have to rely on audio memory - something that has been proven over and over again to be extremely unreliable.  Without test equipment and (experience based) knowledge, you're working in the dark, and have no idea if the valve change has done anything at all.  In many cases, it's actually necessary to modify the amplifier to allow you to measure the current in the valves, and if you don't know what you're doing you can easily end up with an amp that doesn't work, or worse, requires very expensive repairs.  All fixed bias amplifiers should have 10 ohm 5W resistors between the cathode and earth (ground/ chassis), and this makes measurements easy.  The resistors should be matched first, since most are 10% tolerance, and you need better than that.  The resistors should be matched to 1% - the absolute value is not especially important, only that the two (or four, or ...) resistors are within 1% of each other.

Output valves do need to be matched, and this usually requires that you have many more to hand than you'll actually use.  Even if matched sets were purchased (see matching, above), you will still need to verify that they really are matched, and you'll regularly find that supposedly matched valves are quite different.  The first process is to match the valves for quiescent (bias) current, and this usually requires that one valve is deemed the 'reference', and others are graded against it.  Many insertions and removals later, you find a few valves that allow you to select pairs with bias currents that are within 5% (or better) of each other.  Whether you continue the process to the next stage or not depends on the application.

The next stage is to run the previously matched pairs to verify that they have similar gain (gm, or mutual conductance).  Before doing so, you must verify that the grid drive signal is identical for each output valve.  Some hi-fi amps provide a means for adjusting this, but it's rare in a guitar amp.  If the voltages are different, you need to make whatever changes are needed to ensure that they are not only equal voltages, but have the same waveshape.  This typically requires the use of a low distortion audio signal generator set for a frequency of around 400Hz, and a dual trace oscilloscope with the ability to add/subtract one waveform from another.  It's nice to have 100:1 probes because of the high voltages, but 10:1 probes will usually be sufficient.  If the waveforms are identical but inverted, the sum of the two gives a flat line - zero output.  Once the grid drive voltages are correct, you can move on.  Note that if you can't match the output valves perfectly it may be necessary to offset the phase splitter outputs to compensate, but in general it's unwise to do so.

Matching at different power levels is difficult to do in a push-pull power amplifier, because it's very hard to see the possible waveform differences caused by a mismatch.  The 10 ohm cathode resistors come to your aid again.  You can then drive the amplifier to half and full power (using the same oscillator as before), and verify that the voltage measured across the resistors is the same for each valve.  If it's not, the valves aren't matched for gm, so you can now go through the remove/insert (repeat as needed) process again, until you find a pair of valves that is matched for both bias current and equal current at different power levels.  Finally, verify that the amplifier clips symmetrically with the nominal (resistive) load attached.  If it doesn't, that means that either rP (internal plate resistance) is different, which affects the maximum current that each valve can draw, or the valves have screen grids that may not be in perfect alignment, so the valve simply cannot turn on as hard as another.  Well made valves will generally be close enough for all but the most critical applications, but the test is easy to do and may improve your level of confidence.  You'll also get to see the process in action.

note On no account should your oscilloscope probes be connected to the plate connections for the output valves.  The voltages developed are lethal, and can easily damage your probes (even 100:1 types unless rated for at least 5kV) or your oscilloscope.  Accidental finger contact may cause instantaneous cardiac arrest, after which your interest in valve amplifiers will be terminally diminished to nil.

Never underestimate the secretly malevolent intentions of high voltages - they have infinite patience, and are silently waiting with tooth and claw poised to see to it that you don't do it again.  Yes, I'm being frivolous, but 500V DC is not !

Asymmetrical clipping introduces a net DC component into the output transformer (as does imbalance at any power level), which can cause premature core saturation and distortion at low frequencies.  While we all understand that hi-fi amps should not clip, it is pragmatic to expect that it will happen during high level transients.  Whether this is a major issue for you is up to you to decide.  For guitar amps, symmetrical clipping is important, because this is how the amp is operated much of the time.

For a stereo hi-fi amp, the valves in each channel should be matched against those in the other channel, or you may have a channel imbalance that disturbs the stereo image.  Check that normal full power is available - a transformer fault will cause power output to fall dramatically (for example).

You may then decide to run a frequency response test to verify that both channels are the same from 20Hz to 20kHz.  While it's uncommon for valves to have a significant effect on response, it's easy to test while you have everything set up.  If the results of these tests are all as close to identical as you can get, the job is done ... until next time.

See, simple isn't it?  Provided you have the equipment and the knowledge, as well as a stock of valves (which must also be the same brand, type and age of course), the process is pretty straightforward, As described, very, very few enthusiasts will have the necessary equipment.  Before the amp is returned to service, there are a few other tests that should be run.  It's worthwhile to check the ESR (equivalent series resistance) of the filter capacitors.  As these age, ESR increases much faster than capacitance decreases, and high ESR means the caps should be replaced because they are on the way out.

Test that all valve sockets have a good grip on the pins.  Many sockets use rather flimsy contact forks, and these must be re-tensioned before the amp is returned to service.  Also, check for dust or other gunge around the sockets - especially carbon tracks which indicate that there's been a flashover at some stage.  Replace the socket, unless you are confident that you can make a serviceable repair to a carbonised/damaged socket (it's not easy).  A visual check of the amp's interior is needed to spot any possible problem areas that may cause failure at some later date - this requires experience.  If you don't know what to look for, then you don't have enough experience with valve amps.  Any suspect capacitors or darkened resistors need to be replaced, wiring checked, and if any circuitry is mounted on 'elephantide' with rivets as connection points, it should be checked for hygroscopic performance.  Some of this material becomes quite conductive (several Megohms) under humid conditions, and can cause major problems during periods of high humidity.

Finally, the bias adjustment must be rechecked for your newly matched valves in what is now (hopefully) a fully serviceable amplifier.  A final check before closing everything up and you're done.

If you happen to be working on some Fender amp models, you may find that there's a trimpot to allow you to set the bias correctly.  However, (and this is all completely true, verified by the schematic), the valve bias voltage is taken from the trimpot's wiper, and if (when) it goes open-circuit, the output valves have zero bias.  Unless you are right there, with the chassis out of the amp and your finger on the power switch, the output valves will be destroyed.  What should have been included is a resistor from the bias supply to the wiper (probably around 68k), so when the wiper becomes open, the valves get a bit more bias and turn off more, rather than go into full conduction.  This happened to a very experienced guitar amp technician, who didn't realise that there was no failsafe until the worst happened.

This is a very popular Fender model, but it uses the cheapest (and crappiest) trimpot for the most important voltage in the entire amplifier.  The fault is quite common (lots of 'chatter' about it on the Net), but the design hasn't been changed since around 1996.  One would hope that a famous manufacturer would understand the error and fix it, but the amps survive their warranty period so it's obviously not considered essential.

So yes, you can change your own valves, as long as you can do everything above using your own knowledge and experience.  If anything above makes no sense to you, then I suggest that you find an experienced service technician to do the job for you.  If everything looks straightforward, you have the equipment and the background knowledge, and you are confident (but circumspect) around high voltage supplies, then you can replace your own valves, otherwise, DON'T DO IT!


While the above covers the major myths surrounding valves, there are obviously others.  Many of these fall into the 'magic component' category, and I refuse to give these claims any credibility (or web space) by attempting to answer them.  What I've attempted to do here is state the facts, and as readers of my site will be aware, I don't accept the premise that magic exists, so my approach is based on verifiable test methods, and the results obtained from such tests.  The use (or avoidance) of capacitors with brightly coloured or jet black cases because they sound 'better' is just nonsense.  Capacitors are chosen for their value, voltage rating and reliability.  Anything else is nonsense.

If I happen to have omitted your favourite myth, please let me know.  It needs to be able to be verified though - anything based on subjective tests (without the benefit of a double-blind test regime) can neither be verified or refuted, since it is almost certainly imaginary.  The words used by the subjectivists are meaningless, because they don't describe any physical property or measurable phenomenon.  Without these, it's anyone's guess as to what is actually meant, and in any dispute the subjectivist can argue that one simply 'misunderstood' what was said.

There are also several minor myths, including the speed of valves (hint - triodes in particular are s-l-o-w) and silly nonsense such as the points referred to in the intro.  Especially interesting is the inane claim that valves are "better at reproducing deep bass and extended (...) highs".  Sometimes I wonder if the authors of such drivel actually believe the crap they write or if they are having a laugh.


The primary references used for this article were various valve distributors' websites, to see what criteria were used for matching.  Most tell you nothing that's even remotely useful, a few provide some worthwhile info that can be used to verify that the matched valves are likely to perform well, and others say nothing at all.  As in absolutely nothing - not a word about the process, but you may get to see a glowing review (meaningless twaddle) from some magazine or another that waxes lyrical about 'authority', 'immediacy' or perhaps 'intimacy'.


HomeMain Index ValvesValves Index
Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is © 2009.  Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws.  The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference.  Commercial use is prohibited without express written authorisation from Rod Elliott.
Page published and copyright © 09 Nov 2009./ Updated Sep 2014 - minor changes only.