KISS 114
KISS 114
MEASURING PROGRESS BEYOND EXCELLENCE
Andre Jute
It is no big deal designing a standard good 300B amp. You take two stages of 6SN7 (we'll come to them) and a 300B, two of the most linear amplification devices ever made with which only a fool can go wrong, you surround them with carefully calculated values of the best components and you present the whole in a tasty but simple chassis to justify the price you will have to charge for this fine window on the concert room when the amp is matched to the right speakers. This is a class act. It is also a standard engineering job, designed, specified and delimited to work in that part of its operating band and in an audio chain where distortion will not be objectionable. As an SE amp it benefits from the well known psychoacoustic fact that second harmonic distortion is less objectionable than third harmonic distortion, odd harmonic distortion, higher levels of harmonic distortion, of all of which nastiness the standard good 300B has only a very little.
A standard published 6SN7-6SN7-300B design from me (or handful of others) has its little extra justifying its reputation and price provided by careful parts-matching to our wu or artistic sense. Part of that artistic sense is ruthless application of the KISS principle so that the resulting amp has zero negative feedback, no superfluous filters and other excrescences, and of course the minimum parts count to color the sound. Build the amp as we tell you and you can reproduce what we hear, or as near as room effects will allow, without spending years and several fortunes on wasted parts. Change the parts, especially the capacitors, and you can tailor the sound to your taste.
Implicit in these statements is the admission that fidelity has not yet been achieved. There is a school of solid state aficionados who claim that the battle is over. Their measure of success was once inaudible total harmonic distortion (THD). Their amps, complete with provably inaudible THD, still sounded wretched to many of us constant concert goers. Gradually their belief in inaudible THD morphed, much like Marxism when it cannot win an argument on its merits, into a belief that vanishingly small THD is the solution. Many with good ears and a lot of experience with the finest live performers think that amplifiers designed to the THD paradigm, however it may now be stated, still fail to satisfy their perfectly reasonable demand that the amplifier must faithfully reproduce at the ears the experience in the concert hall. We listen with our ears, not with instruments, said a large number of us, including some with decades of engineering experience in audio and elsewhere plus other skills in psychology besides.
Meanwhile, dissatisfied audiophiles started returning to tubes because tubes seem to offer a better representation of the concert hall experience, regardless of their measurements. In fact, it soon came to be believed in a small but influential niche that in tube amps conformity to the ruling audio engineering paradigm of the day (that THD predicts which amps are worthwhile) was an inverse indicator of relative fidelity. For the big powerful, complicated push pull amps with lots of negative feedback (the solid state cure-all) and therefore often superb measurements . . . cast the same chill over the music as silicon amps. By contrast, the worst measuring amps, single ended zero negative feedback amps, which by all the prevailing wisdom should sound like shit . . . in fact pleased most of all.
One can always tell when a prevailing fad is past its best. One merely watches the little men, the hangers-on of the fad, the ones who will never have a new idea of their own, become shriller and shriller in their defense of the most extreme positions of the fad. That is what has happened to THD as the miracle measurement and predictor of sonic bliss. It turned out to predict only which credit card audiophiles are most easily parted from their dosh. The big men already slipped away who the year after next will have the next big idea neatly dressed up in attractive marketing jargon. (I have left out an intermediate stage, in which IMD was, to my mind much more creditably, the big cure-all, and a couple of lesser false starts. If you are reading me, you are smart enough to get the idea without the iterated boredom.) The little hangers-on of audio engineering are screeching ever more hysterically that there is no God but THD, and He is the only God of every audiophile, and we, His Faithful, will kill all the unbelievers, infidels!, who dare enjoy themselves.
Meanwhile, it left me with very little to fall back on except my taste and wu. Having a firm belief in physics as the ultimate explanation of almost everything, I was not too keen to go down the cablie route of opinions apparently based on black magic. There had to be an explanation that could be tested and proven in a laboratory and, arising from the explanation and tests, an engineering measure of success which could double as a predictor of further success.
In fact, what I wanted was already to hand in Olson's work, in which it was found that a higher level of distortion was bearable if it were second harmonic than if it were third, odd, or higher harmonic. This, I argued, is not actually about THD but about the distribution of the harmonics that make up THD, a subtle but very real distinction at your ears.
It turned out to be the engineering lock to glee, a most pleasing paradox.
At this stage I had already by experimenting, listening and conducting double blind tests with real live classical musicians, discovered that it was not SE that held the magic, though it was more obvious in SE and easier to achieve, but Class A1 operation, and ZNFB or very little negative feedback where absolutely essential. The pointers to Class A1 were in fact pointers to a pleasing harmonic distribution because that is what Class A1 excels at, or at least it is easier to arrange in Class A1.
In search of ever smaller THD I had also by this stage developed my HIGH concept, which was simply high voltage, high current, high impedance, all of which singly, and more so in combination, ensure that the signal running along the loadline crosses the transfer curves at the most advantageous angle, often perpendicular to the transfer curves. This is what we shorthand as a flat response. Other successful investigations used shunts of all kinds (silicon, wound, tube) to stabilize every current to as near constant as possible. I ended up with a bigger tube to shunt the current beside every tube carrying the signal, and everywhere else. A venture in heavy current ITs, leading to all-transformer coupled amps, was also successful. Costs and weight spiraled through the ceiling. Experiments in achieving the same effect without the penalties were partially successful and I came away from them with a liking for the sound of the SRPP (which is not, repeat not, a constant current topology).
At this point statistical analysis of my measurements persuaded me that what matters is not how distorted the output is but how the distortion is distributed. Do not go overboard here. My amps were already far less distorted (THD) than those of every other designer because I was and am willing to pay a higher price for silence than anyone else. An example is The KISS Amp, the overarching subject of this major series, where I have deliberately insisted that those who build it must have speakers of high sensitivity, and where you will shortly see me limit output to less than half of what most engineers would say should be available from the amp.
THD must be under the audible limit already before any other benefit of these principles can be realized.
In KISS 115 I publish instructions for a more detailed calculation of distortion than the easy way I recommended before. Do at least read it and make one or two stabs at a working up an example, just to be able to say you did it. Those who are mathematically minded can find in the sources (the RDH describes a few) even more complicated and detailed methods but there won't be much to calculate or measure of odd or higher distortions in The KISS Amp, so for this amp those methods are of academic interest. If you have the math and you adapt The KISS Amp to other tubes than those under consideration by me, you may find more detailed methods of calculating distortion of use.
THE INCOMPETENT AND THE SUBLIME
All of these considerations, experiments, measurements and analyses conducted over a period of a decade came to a head when a hanger-on of a none-too-competent winder, whose wares I refused to specify, challenged me to a design contest. The contest broke down in a flame war after the power stages had been published so the driver stages were never published, which was probably just as well because I have little taste for kicking a dog when he is down.
The challenger published the power tube stage of his own amp, designed by himself and his wannabe-winder buddy or more likely ignorantly adapted by them from a perfectly competent if dull design by an established professional. To make the two designs directly comparable, I detuned the design I was working on to use the same load impedance as the challenger used in his design (but not a transformer from the wretched winder!). The implication is that my amp in real life, when I would load the plate with an even higher impedance, was even further ahead of his amp.
Here is a comparison of the two power stages calculated not by me but by that distinguished RAT, John Byrns, who also gave the two amps their descriptive names. I had no need to calculate because I already knew what a comparison would show.
|
Bob Cs 'Bubbaland'
|
|
Andre's 'Hedonist'
|
|
|
|
|
|
|
|
Grid
|
Plate
|
Grid
|
Plate
|
|
Voltage
|
Current (mA)
|
Voltage
|
Current (mA)
|
|
0.0
|
102
|
0.0
|
136
|
|
-22.3
|
87
|
-23.4
|
119
|
|
-76.0
|
50
|
-80.0
|
80
|
|
-129.7
|
18
|
-136.6
|
44
|
|
-152.0
|
6
|
-160.0
|
30
|
|
|
|
|
|
|
Load Resistance
|
5000
|
Load Resistance
|
5000
|
|
Supply Voltage
|
350
|
Supply Voltage
|
390
|
|
Plate Dissipation
|
17.5
|
Plate Dissipation
|
31.2
|
|
Power Output
|
5.9
|
Power Output
|
7.0
|
|
2nd Harmonic
|
4.13
|
2nd Harmonic
|
2.83
|
|
3rd Harmonic
|
0.82 |
3rd Harmonic
|
0.03 |
Table Copyright © 1998 John Byrns
*If you don't see a table, the raw data is at the bottom of this section to let you make your own table in a word processor.
Note first of all that my Hedonist (thanks, John), though deliberately run high and hot, is still within 80 per cent of maximum dissipation, still a conservative amp. Next note that the power output of my Hedonist is higher than that of the Bubbaland, while distortion in my amp at full power is 2.86 per cent and in the Bubbaland at its lower full power distortion is at least 4.95 per cent and probably well over 5 per cent when the probably substantial higher harmonics are taken into account. The incompetence of the Bubbaland as an SE amp is announced by a third harmonic distortion that reaches for 1 per cent! As a contest this was already such a massacre of the challenger that no one bothered to calculate how much my Hedonist would be ahead at the 1W which is probably beyond the maximum it would ever see in everyday use.
The excuse of the incompetent so-called designers of the Bubbaland 300B was that they took their operating point directly off the tables Western Electric published. There is of course nothing wrong with those tables if you understand the assumptions on which they are given. These two jokers did not. How they could not have heard that their amp was grossly distorted baffled everyone. Perhaps they listened to it only with instruments, perhaps they really have cloth ears.
That is all very entertaining for those who like schadenfreude but what is the real nugget of truth here, the main beef, the fillet for the bourguignonne? Actually, it is a pity that the headline number of second harmonic distortion in the challenger's amp was so bad at 4.13 per cent because it masks the real triumph of the Hedonist, that the distortion for practical purposes consist only of pleasing second harmonic, with third harmonic being simply inaudible at 0.03 per cent, a silicon dreamland number.
It is that tiny little distinction which makes my amps sound so much better.
It is the engineering face of musical ecstasy, both its measure of success when measured and its predictor of future success when calculated on the transfer curves.
In the larger scheme of things then, THD is useless to measure anything you really want to measure because it doesn't predict anything.
What you really want to know from any tube amp with ultra-fi pretensions (DHT SE ZNFB) is:
1. THD at 1 or 2 watt is below audibility which I arbitrarily set at 0.75 per cent because that is where distortion begins to be noticed by the most critical listeners.
2. At any power up to twice or thrice the maximum recommended draw with sensitive speakers, second harmonic should be in excess of 90 per cent of all distortion. A really good amp does much better than the bare minimum on this parameter. (Mine in the contest above managed 99 per cent at full power and would have done better still with the 5K6 tranny I specced for the real amp.)
3. Even at full power third harmonic distortion is tiny, many, many decibels below the fundamental, below second harmonic, below audibility. In other words, the designer disappeared it.
Those in the habit of making semantic arguments will notice that the third point, the absence of third harmonic distortion, is a corollary of the second point, that virtually all residual distortion should be second harmonic. I think it is worth making both checks, writing down both numbers on a sheet of paper and to sit studying them for a while. I usually play one of the more triumphant Bach cantatas or soaring masses while I do it.
Iterations, damned iterations
Amplifier design doesn't in real life proceed in the tidy steps I showed you in designing the standard good 300B output stage. Experience, desire and your junk box put bits of the amp in place out of order and you know before you reach, for instance, the power supply, that you will have to redesign one or more stages because the supply won't make the voltage at the current required. That is what happened here right at the power tube stage.
The difference between the two amps we are designing side by side, standard good and ultrafi, is much more than just different drivers. The standard good amp can have its power rectified by a silicon bridge or individual silicon diodes to taste; the silicon bridge makes a small contribution to the silence of its blameless character. But the ultrafi amp, as we shall see again and again, is shaped by deliberate hedonism, and quite a few other considerations, including history, which at first glance has nothing to do with engineering and everything to do with taste. (We shall discuss the generality of these considerations later, when the history becomes pressingly pertinent, in the section on wu.) One of these matters of taste is using a tube rectifier because it adds a certain sonic signature, or removes it, to choice; by the choice among the available suitable rectifiers, which actually number quite as many as suitable output tubes for this class of amp, the designer gains a small amount of control over the quality of the glass in his window on the concert room. Eventually all these small amounts of control applied in the correct proportions add up to bliss.
KISS demands a stereo integrated amplifier, everything on one chassis with only one power supply and one voltage gain stage which will double as the driver stage. Besides reducing the bulk and the real estate demands of the amp, another design parameter is that I want to use a matching power supply transformer to the Lundahl output transformers because I have good experience with it and have one sitting on my bench already. I know from experience that a choke input filter is another soundshaping element in the amp, and an important contributor to silence as compared with a capacitor input filter (often called a pi filter for its characteristic shape on schematics). We must therefore work within voltage available from the chosen power transformer after tube rectification and after the choke input filter.
In a process I shall show you later when we design the power supply, I determined that what would be available for the 300B was about 396V and 65mA each, which after the copper drop of about 11V in the primary of the OPT would be 385V delivered to the plate, split as -65V negative grid bias and 320V quiescent plate voltage. The same 1K cathode resistor and 51uF cap as for the standard good 300B would do well. Output power would decline a little and distortion would increase but I didn't bother to calculate either because I have no intention of using a 3K primary on a 300B to drive 100dB horns requiring only a fraction of watt.
The back-and-forth iteration of discovering that in an already-chosen power tranny you don't have the juice is another difficulty you have to overcome quite regularly unless you are rich and can just order up custom transformers or unless you wind your own. There is a way of making precisely the necessary voltage from a near-enough power transformer you already own (or can order from the same place as your chosen output transformers), which we will discuss in the chapter on power supplies.
The output impedance
Now, having put our minds in gear to discover what really matters, we are in a position to continue The KISS Amp, where we just decided the power stage should at zero signal stand at 320V operating, -65V negative bias, 65mA on the plate.
If you're looking hungrily at the 80mA, -80V and 390 operating volts of my Hedonist above, don't get carried away. I have no intention of running my last pair of matched gennie Western Electric 300B that high and if you're a newbie, even if you are working with relatively inexpensive Chinese tubes, you might want to cool your ardour until you have more experience. That amp was designed for a fellow who will never again have to count the cost of a pair of WE300B.
We also inherited from the standard good 300B amp a 3K primary impedance hookup on the OPT loading the plate. 3K was chosen to max out power on that very silent amp. But in fact, with either the
Lowther horn speaker or The Impresario design I have published as a more economical alternative, 7W is gilding the lily. We have long since decided that all you really need is two or three watts of very clean power.
We cannot do anything about the quiescent current without changing our power supply transformer or our desired supply filter principle, both of which I am loath to do. We will return to the power supply after designing the driver stage of The KISS Amp.
The only other element still within our control in the output stage is the primary impedance setting of the OPT, which I remind you is a Lundahl 1623-SE 90mA. This very sophisticated cut-C-core tranny is so segmented that its secondary can be configured to reflect an impedance of 5K6 onto the primary and thus the plate of the 300B, instead of the 3K with which we have been working until now. We make new links on the transformer already in place and voila! 5K6 will flatten the transfer function much more than adequately to give us a very quiet amp with an agreeable harmonic distribution. The price in power may bother engineers (it may bother them a lot if they know the dollar price of gennie Western Electric 300B!) but we don't care. The 3.8W The KISS Amp when finished will produce is still higher than the rather generous outermost maximum of 3W we calculated.
With our 100dB horn that is enough drive for 106dB or, more sensibly, to allow us to trade in some of the signal swing for silence in the driver stage, which in turn will also linearize the 300B further.
So now you know the purpose of the transfer curve set
113KISS300Boperatingpoints.jpg
that you've been looking at. It is a point I arrived at after some thought, this point in fact. In a few simple lines it shows virtually all the trade-offs and compromises made in the design of a 300B power stage and how these compromises and trade-offs will control the rest of the design and the sonic signature of the amp. To the thoughtful there is a cosmos of information, meaning, implication and opportunity in those few lines on your screen: that is why I (and I'm not the first) say the sheet Ia-Eg-Eb curves is your most powerful design tool.
*Raw data for table in body of section
Copy the text below and use a word processor to make a table, setting 4 columns, 16 rows, comma separator. In Microsoft Word on the Mac, the procedure that works (the 'undocumented feature' aka workaround to the bug) is to select the table text, pull down the menu Table>Convert>Convert Text to Table, click the 'comma separator' button in the dialogue box, see 5 columns 16 rows automatically selected, and to press return, after which you can manually select and delete the fifth, empty column. Don't shoot the piano player, he's playing the music on the sheet! (Table (c) 1998 John Byrns)
Bob Cs 'Bubbaland', , Andre's 'Hedonist', ,
Grid, Plate, Grid, Plate,
Voltage, Current (mA), Voltage, Current (mA),
0.0, 102, 0.0, 136,
-22.3, 87, -23.4, 119,
-76.0, 50, -80.0, 80,
-129.7, 18, -136.6, 44,
-152.0, 6, -160.0, 30,
Load Resistance, 5000, Load Resistance, 5000,
Supply Voltage, 350, Supply Voltage, 390,
Plate Dissipation, 17.5, Plate Dissipation, 31.2,
Power Output, 5.9, Power Output, 7.0,
2nd Harmonic, 4.13, 2nd Harmonic, 2.83,
3rd Harmonic, -0.82, 3rd Harmonic, -0.03,
THE VOLTAGES IN THIS AMP WILL KILL YOU.
GET EXPERIENCED SUPERVISION IF IT IS YOUR FIRST TUBE AMP
All text and illustration is Copyright © Andre Jute 1998, 2004, 2005
and may not be reproduced except in the thread KISS xxx on rec.audio.tubes
