I can guarantee you, the frequency response was as flat as a ruler with all tested ecc81 variants.
Don't get me wrong, I would very much like/want measurements to correspondent to percieved sound. Alas, in this case it wasn't. I do not have an explanation for it.
Thanks for sharing your test results, the simple explanation might be that the differences were due to the tube's tolerance, i.e., the tubes' constants (gm, rp) were not identical, so unless you spent the time & money to match both sections of all the tubes to begin with, why would they sound the same?
I agree completely with you; however that is not exactly what I meant.
What I can't explain is that harmonic spectrum and frequency band are more or less the same, but the sound is not. I expected more differences in spectrum between different tubes. And I expected correlation between the spectrum and sound.
The frequency spectra taken with a 1 kHz input are close, but what about other frequencies? What happens with harmonics of 5 kHz? Or IM?
Sorry, didn't try that. If I find the time (and motivation😀) I will try that as well. I don't think it will change much, but you never know.
Different valves have different distortion spectra, but near enough identical frequency responses. Different valves sound different, exactly as you would expect from different distortion spectra. I can't understand why anyone might think otherwise.
Maybe some of us are underestimating our abilities to hear distortion.. Maybe that is the magic some people think they can hear that can't be measured.. It really is that tiny tiny measurable difference in distortion that makes a larger impact on what we hear, than we'd expect.
I'm questioning "If there is a difference in perceived sound, then what is responsible for that difference".
Is it the minor difference in distortion spectra?
Is it the minor difference in IM distortion?
Is it the minor difference in floor noise level?
Is it the 20% tolerance in parameters such as Gm and Rp?
If there is a difference in perceived sound, then it follows that there has to be a difference in parametric behavior which is responsible for the that difference.
I suspect that is the case; but a little bit different than "standard" is assumed.
I personally heard some experiments with a distortion generator. (for hifi, not guitar)
With this you could dial in distortion. Different harmonics and percentage.
It was a room filled with about 30 people, so results where a little bit hard to hear. Nonetheless I couldn't hear 2nd harmonic below 1%. And even above 2% didn't sound bad.
The unexpected was 3th harmonic. I always read it sounds bad, but even here 1% was not easily heard, and around 2% it resulted in bright sound.
4th harmonic was terrible already below 1%. Sound immediately became harsh. I can't remember if higher harmonics were tested.
There seems to be thruth in the stating that higher harmonics are "worse" than lower harmonics. This makes it harder to measure. IF the higher (5th and higher) are responsible for the differences in sound, than it is almost immeasurable, as these higher harmonics are below -110 db or even lower.
All good questions😀
And I totally agree with your statement. I just don't know what it is.
I can only guess.
I suspect it is not the noise floor level or the Gm and Rp.
Could it be expectation bias?
Otherwise, I have no idea what's going on here. Including NFB should make performance less dependent on active/passive components.
No, it is not expectation bias😀...this time...
You know, somethimes you change things in an amplifier and wonder if its better or worse, and think it is better but don't know for shure. This was not one of those times. It was a clearly audible difference, heard by 3 men. I bet I could hear which tube was plugged in from another room🙂.
Maybe it is IMD like suggested. I will measure that eventually. I would very much want to know what is going on.
O, and the amplifier uses no feedback.
You know, somethimes you change things in an amplifier and wonder if its better or worse, and think it is better but don't know for shure. This was not one of those times. It was a clearly audible difference, heard by 3 men. I bet I could hear which tube was plugged in from another room🙂.
Maybe it is IMD like suggested. I will measure that eventually. I would very much want to know what is going on.
O, and the amplifier uses no feedback.
It's so difficult to be certain of avoiding expectation bias - and it can readily be a contagious experience, so even second opinions aren't necessarily conclusive IME. When measurements and audible results don't tally, it's only fair to scrutinise both in equal parts.
A mistake which people often make is that if something is good/bad then more of it must be better/worse. Sometimes that is true, but not always. So if objective sound component A makes the sound have subjective quality X it doesn't necessarily follow that having 2A must mean 2X too. Conversely, that fact that 2A does not necessarily imply 2X does not mean that A does not mean X.
So, if a little 4th harmonic make the sound bright/harsh it does not necessarily follow that more 4th harmonic will make it brighter/harsher. At the very least it depends on the levels of other harmonics too, due to masking. So the fact that lots of 4th sounds better does not mean that we can conclude that a little 4th is not responsible for a little harshness.
What we can say is that we would expect different electrical signals (either in the time or frequency domain) to possibly sound different, and that a sound difference must be caused by a difference in the electrical signal. The most obvious one, known from experiment, is that changes in the Fourier component amplitudes are often easily audible, even though they may have only a small effect on the time domain waveform. Changes in Fourier component phase are much less audible, even though they can have a huge effect on the time domain waveform.
Swapping valves in an amp with little or no feedback can be audible. This can be because of the large variation in individual samples off the same production line, or differences in design from different production lines. They will have different bias, different gain, different stray capacitance etc. Which one sounds 'better' or measures better depends on the circuit details as much as the valve. As a general rule, the bigger the difference between samples the worse the circuit design.
So, if a little 4th harmonic make the sound bright/harsh it does not necessarily follow that more 4th harmonic will make it brighter/harsher. At the very least it depends on the levels of other harmonics too, due to masking. So the fact that lots of 4th sounds better does not mean that we can conclude that a little 4th is not responsible for a little harshness.
What we can say is that we would expect different electrical signals (either in the time or frequency domain) to possibly sound different, and that a sound difference must be caused by a difference in the electrical signal. The most obvious one, known from experiment, is that changes in the Fourier component amplitudes are often easily audible, even though they may have only a small effect on the time domain waveform. Changes in Fourier component phase are much less audible, even though they can have a huge effect on the time domain waveform.
Swapping valves in an amp with little or no feedback can be audible. This can be because of the large variation in individual samples off the same production line, or differences in design from different production lines. They will have different bias, different gain, different stray capacitance etc. Which one sounds 'better' or measures better depends on the circuit details as much as the valve. As a general rule, the bigger the difference between samples the worse the circuit design.
Well, then let me ask one question:
Why don't all electrical /sound engineers have good old chip amps from the 80's/90's?
Those where very high feedback, opamp alike amplifiers with near perfect THD and IMD; often less than 0,03%.
FFT looks very good from these. In fact everything we normally measure to determine if an amplifier is "good" is near perfect with those.
But even if I got them for free, I wouldn't listen to them. (anymore)
And I'm pretty shure thats the same for a lot of you. Otherwise you wouldnt be interested in a tube forum....
And i'm pretty shure it has nothing to do with "psycho acoustics"
I'm also shure it is not the higher distortion of tube amplifiers wich make them sound good. Some tube amps sound better that other, and mostly I like the "simple" designs with low distortion the most, sonically.
Just my opinion: Measurements are not some holy grail, but an aid. There are important aspects to amplifiers wich are very hard to measure and to express in numbers or percents.
Why don't all electrical /sound engineers have good old chip amps from the 80's/90's?
Those where very high feedback, opamp alike amplifiers with near perfect THD and IMD; often less than 0,03%.
FFT looks very good from these. In fact everything we normally measure to determine if an amplifier is "good" is near perfect with those.
But even if I got them for free, I wouldn't listen to them. (anymore)
And I'm pretty shure thats the same for a lot of you. Otherwise you wouldnt be interested in a tube forum....
And i'm pretty shure it has nothing to do with "psycho acoustics"
I'm also shure it is not the higher distortion of tube amplifiers wich make them sound good. Some tube amps sound better that other, and mostly I like the "simple" designs with low distortion the most, sonically.
Just my opinion: Measurements are not some holy grail, but an aid. There are important aspects to amplifiers wich are very hard to measure and to express in numbers or percents.
You are jumping to a number of conclusions.
Everything who measures? Reasonably good THD and IMD figures are good things to have;a mathematician would say that they are necessary but not sufficient. Once they are good enough, making them better does not improve sound.pauldune said:
Why? Some may have sound faults which don't show in THD/IMD measurement. However, some may simply produce excellent sound but this is not what all people prefer - some like some distortion or lumpy frequency response.
You (probably) have no idea why we are here. Some like 'tube sound'. Some hate it, but still use valves (for various reasons).
It is very likely that it is the higher distortion of some tube amplifiers which make them sound 'good' to some people. This is clear from the popularity of 'tube buffers' - how can adding noise and distortion to a signal improve it unless some people prefer noise and distortion?
What are these "important aspects"? If you know what it is then you ought to be able to measure it. If you don't know what it is then how do you know it is relevant?
Then we, of course, need to define "worse" and the context.
If our goal is eliminating all distortion then all harmonic distortion should be considered bad.
If our goal is to make an "effect processor" to produce "euphonic distortion" - or whatever people wish to call it - then we may have individual preferences concerning what harmonics are good and in what specific application. Some may prefer "warm" sound of moderate amounts of low order harmonic distortion. Some may find it too subtle and prefer lots of high order harmonic distortion in addition (e.g. effect processing for pretty much all modern guitar music).
But then our discussion quickly starts to wander off to subjective realm instead of objectivity.
This. I read between the lines the spectrum plots are taken from speaker terminals of a 1W amplifier. That's a complete system where several different variables could affect the test results and setup.
For example, the loading characteristics most likely are not purely resistive and may change at different frequencies. Within tube parameter spread one tube might handle it differently than another. This would undoubtedly create a different spectra.
Last edited:
- Status
- Not open for further replies.