|
TUBES vs.TRANSISTORS PART II by Steve Deckert Since this email is so lengthy, I am going to respond to it a point at a time. My responses will be in bold face. Steve, Thank you for your detailed and thoughtful reply. However, I must take issue with you on a number of things you brought up. First of all, I don't listen to music (mostly classical) with my amplifiers operating at a nominal level of 10 watts. The average listening level is closer to 1 watt, which I have measured numerous times with my HP A.C. voltmeter and "Mitey-Mike" based sound level meter. Besides, if your assertion were true that 10 watts is closer to nominal, then how could your Zen Amplifier, Pass's Zen Amplifier, or any of the other under 10 watt Class-A Single-ended triode amplifier possibly perform without producing excessive distortion at more than moderate levels? You are responding to the following statement that I wrote in my previous email: "using your 200 watt amplifier as an example and listening at a nominal level of 10 watts, you would need 1280 watts to reproduce a 20 dB musical peek without clipping the wave form." I don't listen to my system at an average level of 10 watts either, but then I didn't say that was the average listening level, I said at a nominal level of... thinking at the time what it would take to reproduce live music at live levels on Ave. efficiency speakers of 86 dB. Make no mistake, solid state cost effectively gives more people the option to play music loud than tubes. As far as average listening levels, I also agree there is more content in the first magical watt than most people realize. The Zen amp on the same ave. efficiency of 86 dB speakers cannot perform without producing excessive distortion at those levels, nor could (as you point out) any other 10 watt amplifier. The way this is compensated for is by using more efficient speakers. If you wanted to achieve those levels you could easily do so with large horn loaded designs. My home system is tri-amplified, which I think is the only way to go for true "high-end" performance. The output of my power amplifiers (7 channels, including an extra sub-bass servo-feedback sub-woofer (my design) in the rear of the room) drive my speakers directly, with no passive crossover to attenuate (or otherwise "color") the signal. My power amplifiers (custom-modified MOSFET -based kits) have massive power supplies--8 amp. toroidal transformers and 80,000 mfd. of power-supply capacitors per stereo-pair. These amplifiers drive the most excellent Morel MW-142 woofer and MDT-33 tweeter in small satellite enclosures, which I maintain provide better imaging and lower coloration than most (if not all) of the larger system designs (which, often have four and five figure price tags). You probably are not all that impressed with measurements, but, using a pink noise source in 1/3 octave bands, my system measures within +/- 3 dB from 40 Hz. to 18,000 Khz., on axis, or 30 degrees off horizontal axis. This is partly due to the very small cross-sectional dimensions of the midrange/midbass driver, the low mid/high crossover frequency of 1,900 Hz., and the egg-shell-like small enclosure design. I hope to perform pulse-response measurements soon, using a spectrum analyzer program on my computer. Actually I am impressed by measurements (specs), they are the only tangible non subjective thing we have to analyze performance. I just don't care what the specs are if something sounds good and rather like not knowing what the specs of a particular piece of audio gear are before I listen to it. I like to audition audio gear with an open mind, and the less I know about what I'm hearing, the more I can hear. With the reserve instantaneous available power provided by my power amps, at even high levels for home-listening in my 12 X 20 ft. listening room (which, by the way has been acoustically optimized for the system, and the satellites and sub-woofers have been optimally positioned--away from the wall, etc.) I can achieve more than adequately high listening levels without audible clipping. Audible clipping - I assume that means a form of distortion that you are consciously able to detect. At high levels clipping could be harder to consciously detect against the 100 dB "noise floor" created by the music. However, I'm not saying YOUR system is clipping. Let's focus on clipping a moment. A good solid-state amplifier will clip a sine wave coming from a test generator with absolute symmetry--perfectly flat on the top with no droop, and absolutely no overshoot or spikes. This was the problem with the earlier transistor designs and caused them to sound harsh. I followed this carefully back in my college days (late 60's) when I was selling and installing the new Marantz 8B's and Model 9's to the more wealthy people in my community when I was working my way through college. Yes, my first amplifiers were tube-types, Dynaco Stereo 70's and Mark III's, and they sounded damn good, I almost cried when I finally got rid of them. This was when the new solid-state units were just coming into existence (Carver, et al.). Let's also indicate that when a solid state output device clips it is actually shutting off and then turning back on again creating DC on the voice coils of your speakers, and causing the speaker to stop momentarily. Severe clipping means not enough air moves over the voice coil and it over heats, the number one cause of blown speakers. This is why it is possible to blow up a 100 watt speaker with a 10 watt solid state amplifier. This sort of well designed transistor power amplifier is capable of amplifying the dynamics of music without audible clipping. It has been my careful study throughout the years that the only clipping that occurs, does so only at the relatively fast transient peaks, and because these peaks are fast, they simply cannot be heard. Why? Because these harmonics occur for only a brief few milliseconds and the psychophysics of the human hearing mechanism simply does not have the neurological apparatus to detect them--the way the cochlea and its associated neuro-waystations simply mask these effects is well understood. Read a good book on psychoacoustics to find this out. Frankly I could write a good book on psycho acoustics. It is my main focus. What happens in the mind to make one think a system sounds good, is the perspective from which I design audio gear and the very reason I can come across as having little respect for specs. And you are right about our inability to consciously hear the relatively fast transient peaks. Again if you like numbers the neurons in the human brain can not fire fast enough to distinguish phase above around 2800 cycles, yet the effects of such can be heard in the way they effect the sound stage. The simple equation dB = 10 log (P1/P2) Would then indicate that if at my nominal 1 watt for an 85 dBm or so acoustic level, then the loud passages of the orchestra are allowed to obtain over 110 dBm of acoustic level before my amplifier would begin to clip to a point that could be audible. Remember, my amplifier can produce 300 watts or more for a several milliseconds without clipping--long enough to sustain nearly all of these short duration peaks without clipping. That's what is called reserve power, something that just is not present with your preferred low power Class-A single ended tube designs--all they can do is squash these peaks severely and in so doing throw in the signal all kinds of harmonic content (coloration) that simply is not supposed to be there--if you want reproduction that is close to the original source. It's ironic that you criticize my 200 watt RMS amplifier based upon its insufficient power and yet advocate the use of a 6+ watt single-ended Class-A amplifier in you Zen Amplifier article. I suspect that such an amplifier would be very effective in driving headphones, but not much else. I don't think there is an argument here about clipping, I simply think tubes can sound better. I am sure your system is not clipping 99% of the time at the average listening levels you operate it in. Reserve power, called "head room" is a monumentally important thing to have a lot of in a good solid state amplifier for reasons we have both pointed out. If solid state amplifiers didn't clip so aggressively (flat line DC at the peaks) the amount of head room would not be a serious issue to the listener. In a tube, clipping is completely different. Since a tube works by passing electrons from a cathode through a charged screen to a plate by way of different electrical potentials, there is no such thing as "clipping" as we've come to know it through this discussion. The tube never shuts off, and never puts DC on the voice coil. It never creates an unnatural flat line at the top of musical peaks. What happens is more like a bucket (the plate) being filled with water (electrons). When the plate is saturated there is no longer a potential difference - so, no more electrons are accepted on the plate until there is room made on the plate by the dissipation of those electrons. Instead of clipping, a form of compression occurs as the difference between the continuous power and peak power are reduced. I too have spent many hours observing amplifiers clip on an oscilloscope. In fact I have a special oscilloscope (formally used for medical signals) that has a high-persistence (P7) phosphor, which is ideal for catching the peaks of musical content. I have also built a special sample-hold circuit from an LED dB level meter I.C. which has been carefully adjusted to light an LED and hold it on for a half-second any time my amplifier clips (I also have to compensate for changing power line voltages). Using this apparatus, I have found that my amplifiers rarely clip. And, when they do, it is of such short duration that it simply is not audible. Incidentally, as an experiment, I once connected a very fine 30 watt solid-state amplifier to my Morel Woofers and found it to be inadequate--its clipping was indeed audible! Thank god it did not disguise this clipping by squashing it! This was a clear indication that I needed more reserve power. ...or perhaps a 30 watt tube amp. And yes, thank god it did not disguise this clipping by compressing it because that would have been musical by comparison. OK, what about these clipped peaks? Observing this on the scope, these peaks are again, relatively brief. A properly designed amplifier of adequate power will clip these peaks (which, remember, occur rather infrequently in the context of the rest of the mass of the musical content) cleanly, without overshoot, etc. Thus, the harmonic structure of this clipping is of very high frequency and low power content. It is well above 3,000 Hz., and probably exists mostly as harmonics above the 18 Khz. audible top end--remember, these are harmonics of already very brief peaks--peaks of generally less than 5 ms. which are too short in duration to be perceived. My observations have been that solid state clipping can happen anywhere from 20 Hz to 20 kHz and usually does. In fact the kick drum centered anywhere from 40 to 100 cycles depending on how it's tuned, can usually by found on a scope as the number one clipping frequency. I didn't know you could have harmonics of a fundamental clip that high up the spectrum without either the fundamental itself or one the first three harmonics of that fundamental clipping. I also need to mention that the load given to the amplifier by the speaker at any given frequency and the amount of negative feedback will largely determine where in the frequency band an amplifier is most likely to clip. To think of an amplifier as a separate entity would be misleading since the speaker completes the circuit, and no two speakers are alike. It has been demonstrated that the perception of "harshness" occurs in the critical 800 Hz. to 3 Khz. region, and it has to be sustained long enough to be perceived by the hearing mechanism. The distortion produced by the occasional instantaneous clipping of a sharp music spike in an amplifier has most if not all of its power-spectral content well above this critical frequency range. Yes, on a linear scale it might, but hearing is non linear. Our ears have the highest sensitivity in the 800 Hz to 3 kHz region, so even though the majority of clipping may occur outside this band, any clipping in this band would be perceived. You say that "even order harmonic distortion of tubes can be a non-factor before clipping in a good circuit". That's basically true. A well designed tube amp will probably have less than 0.1% harmonic distortion below clipping, which is inaudible. But the problem comes when the tube-amp runs out of steam and begins to go into clipping. I find 200 watt (RMS) tube amps. insanely expensive, and almost non-existent. The coloration of a tube amp. is more "consistent" (your words) if it is lower in power and thus able to squash the signal more readily. Well designed 200+ watt RMS solid-state amps. are readily available, easy to maintain, fairly power efficient (class A-B), and not too expensive. The debate is not about practicality, but musicality. Tube amps a not terribly practical. In fact a good one is not unlike a pet, requiring special attention and upkeep. Solid state was a less expensive more reliable means and largely more profitable way to build amplifiers so everyone jumped on it. The
differences between the even order harmonic distortion
So, now comes the High-End (Retro Ghetto?) Politically-correct single-ended Class-A Triode amplifier. Generally, these products can produce about 20 watts maximum. So just imagine how these "waveform squashers" color the signal with all but perhaps the most efficient loaded-horn type of speaker systems (the latter of which really are "colorizers"!--believe me, I spent dozens of hours in listening rooms during my college days listening to these beasts, the only thing worst was the horrible Bose 901). So, Steve, in your Zen Amplifier article, you say that amplifiers contribute more to the quality of the sound than anything else, including speakers. Yes, for these "Class-A Puny-Watt Triode Wonders", you are probably right. They should definitely have the ability to muck up the sound more effectively than anything else in the signal chain. Living
in this country and market, I clearly expected you would also be
an anti-horn person, so let me just say that the horns you listened
to sucked, and you probably heard them on solid state gear which
is a no no. Good horn speakers have the most intimate coupling
to amplifiers of any speaker. For that reason, you do not
want to try to separate the two with feedback, and that eliminates
almost all solid state circuits. There are no good horn
speakers in this country. What we really want is zero distortion, but this is audio where no one can have it all. And IF we ran amplifiers to a level where they begin to distort, even order would be far less of an offense that odd order would it not? You imply again that tubes are always producing even order distortion, yet you yourself agreed that a good tube circuit operated within its parameters can have less than .1% I have been told and have done double blind listening tests to see, that humans can't detect harmonic distortion until it reaches 3%. I found that to be basically true, so when I see people basing buying decisions on the difference between .01 and .0001 % harmonic distortion I just have to shake my head. Because of the fewer parts needed to construct a tube circuit, I find a good tube amp not only has LESS coloration than solid state, but it can also sound faster. What I want to achieve is a music system that does nothing to the signal coming from the source--be it compact disk or master tape. I want all the creativity, artistry, and hall effects to end where the rarefactions of air hit the microphone capsule. At least as much as possible--I know I am still victim of the whims of the recording engineer. But the recording companies are getting better and better all the time. A fine example of this is the improved CD quality of Deutsche Grammophon over the years. The classic statement made by all engineers and audiophiles absorbed with specifications.. I want the same thing stated in a different way... to achieve a music system that breaths with openness and is not veiled by negative feedback and excessive parts to insure great specs. You see even more important than great specs are the things that specs can not yet measure, like clarity, depth, width, focus, delineation. These are the things that determine how REAL a recording sounds. If you take your average solid state amplifier with perfect specs and compare it with my Zen amp, you will discover that the Zen amp lets you hear several additional levels of detail, the result being far more realistic and involving at moderate listening levels. Unfortunately phonograph records and vacuum tube amplifiers don't do this, they add color--something which was not present in the original. After listening to such a system for months or years it would be natural for one to say that a CD/Solid-state system sounds thin or lacking in warmth, etc. But on this basis so would perhaps be the sound of a live performance. This reminds me of the psychology experiment where young chicks are given two containers of "water". In one container is pure water, but the other container contains mercury. the chicks immediately go for the container of mercury--it is more shiny and reflective than the water--a sort of super-stimulus--more warmth, more air around the instruments. In college I got the same effect smoking a joint! I can't agree with that, but I will suggest that getting stoned before listening would be similar to hearing my Zen amp. You would be so lost in the music you wouldn't care about clipping. I wonder what would happen if individuals of the Pro-Tube camp were asked to listen to two sources hidden behind a curtain. One curtain would be a live performance--say a string quartet playing Schubert, while being miked. And, they would then be asked to listen to the amplified reproduction of that string quartet through the best Single Ended Triode Class-A amplifier system money can buy. I suppose they would favor the latter, as they would hear the warmer sound and the increased "air around the instruments" produced from the increased coloration introduced by the amplifiers. They would favor the later because the later would have the ability to deliver a far more accurate presentation. In fact it should be almost impossible to tell the two apart. I can assure you that the triode camp could pin point a solid state amp in the signal path in less than 5 seconds every time. And, your comment about negative-feedback. You, and many others like to imply that negative feedback is a bad thing. Sorry, but this only indicates your ignorance of electrical engineering concepts. You should refer to articles already written by Tom Nousaine, David Rich, Tomlinson Holman and others who have already gone into depth about such misunderstandings. Negative feedback is part of the reason solid state amplifiers can achieve extremely low distortion figures of say less than 0.05%. But, according the Pro-Tube Camp, low distortion must be a bad thing--something to be avoided, or at least not discussed. It's no wonder then why manufacturers of tube amplifiers seldom publish any meaningful distortion figures. Besides, according to many who make all those subjective testimonials, careful laboratory measurements are meaningless. If what they claim is true, the laws of physics somehow magically do not apply to audio equipment! The laws of physics do not apply to the laws of psychoacoustics, nor do specs (as I previously pointed out) define all the variables of realistic playback. Also I do not feel terribly ignorant of electrical engineering concepts. I obviously have done more with negative feedback study than yourself because I was willing to look beyond specs. Oh believe me, negative feedback is the best thing since sliced bread for engineers. It lowers distortion and raises dampening by a considerable amount. The problem is that by taking the output of an amplifier and running it back into the input stage, you create a time delayed input super imposed over the original signal that veils and homogenizes the output. Quite frankly it is a speaker dependent argument, but in general I find negative feedback has the exact same sonic effect as throwing a blanket over your speakers. Then you mention that solid-state amplifiers have more capacitors in the signal path. Well, nothing could be further than the truth. Besides, a properly spec'ed coupling capacitor has absolutely NO effect on the sound. All it is doing is blocking DC. All of this talk about one brand of capacitor having superior sonic characteristics to another is absurd superstition and could never be proven with objective A-B listening tests. Besides, most modern solid state power amplifiers have only one coupling capacitor. It is at the input--to block any small DC offsets. The rest of the design is direct-coupled, all the way to the loudspeaker. My home system has no passive crossover network at the loudspeaker either. I'm using 24dB/Octave (Linkwitz-Riley) active equalization using 1% matched components for nearly identical response characteristics from both stereo channels for maximum "holographic" effect. The instruments of my reproduced orchestra have excellent spatial placement. In fact, better than anything I have yet heard at my local High-End Saloon listening to five figure-systems, but, of course, this is just my subjective (and therefore biased) opinion. Here, Double-blind ABX testing would be a bit difficult. Your
position on coupling capacitors is depressing and absurd.
Looking at the sketch of your Zen Amplifier at the top of your article, I believe I see three coupling capacitors per channel, if this is true, aren't you being a bit hypocritical here? The vacuum tubes and output transformer are much bigger contributors to coloring (distorting) the signal than a coupling capacitor. A coupling capacitor simply exhibits a very low reactance (resistance) throughout the entire audio range--that's all. And, all the absurd theories by proposed certain tweako capacitor manufacturers are just plain fancy. One 2 Mfd. mylar capacitor will have the same effect as any other 2 Mfd. mylar capacitor--it has no "sound" of its own. The
schematic you're referring to was an integrated amplifier from which
the Zen amp project was born. The Zen amp is only the last
stages of that circuit, and does use only one coupling cap.
And none is used on the input, the input is direct coupled.
I stuck with my Dynaco Mark III's up until the early-eighties, for my more serious listening of opera and orchestral music. But, as a hobbiest, I began to build better and better solid-state amplifiers. And, when the better MOSFET designs came out, it was clear to me that it was time to make the switch. No more annual replacement of those expensive 6550 matched pairs, or monthly checking of the grid-bias. Better, cleaner, purer sound resulted from my transition to solid state amplifiers. The Apt/Holman company nearly 20 years ago designed a transistor amplifier that would simulate the squashing effects of a tube-amplifier going into clipping. Using special circuitry they made the edges of the clipping rounded, just like the effect you get with vacuum tubes. They then did some serious blind-testing with a group of listeners, comparing this experimental design with their standard design which clips very squarely, with no overshoot or droop. And, to their amazement, their listeners reported that they preferred the sound of the un-modified square-clipping amplifier to the sound of the vacuum-tube-like clipping effect. I wish I had my hands on the paper they published on this--I would send it to you. Yes, I should think they would have found the unmodified amp more open sounding , and the other modified amp to be slightly more veiled by the additional complexity of the circuit. I find the preference much of the High-End community has (as exemplified by Stereophile Magazine) towards the preference to vinyl recordings to the superior CD, and Vacuum Tubes to transistors (especially the Single-ended Class A Triode four and five figure nicknacks probably only purchased by the very rich who, not caring a diddily about the sound quality, only want the stuff to impress their colleagues at cocktail parties) took two equally giant steps backward in the progress of making sound reproduction more real. When these such digressions became rampant, many very brilliant electrical engineers, being designers of fine sound equipment, left the industry to seek other occupations in computer technology, video, etc. I personally know at two of them. They simply could not stand the insanity. A
lot of truth in that statement. No offense Larry, but I am focused in the correct direction. My work will make it possible for many people to enjoy listening to music again who otherwise may not have. Lastly, I find interesting parallels between the High-End audio industry and the fashion industry. Insanely expensive stuff--wives thumb through fashion magazines that advertise four-figure designs while their husbands thumb through Stereophile Magazine imagining themselves owning that five-figure amplifier or speaker system advertised. Oh, yes, tubes are definitely more photogenic than transistors. You're correct about the similarities from a market perspective, I have no real argument with that. Well, Larry, you certainly haven't left anything out.. except possibly an open mind. Let me conclude this, my last response, with a story: Several years ago I had a friend who I used to involve in my speaker design work because he had good ears. He went to school to become an engineer. As the first year passed I watched the magic of audio leave his spirit as his schooling defined boundaries from which to think inside. Once he knew the laws, there was no point in trying to work outside those laws "because it won't work". One day at the repair bench we were finishing up a job, and having a beer when I bet him I could turn the beer can into a speaker. I explained that by cutting it in half and installing a magnet inside, reassembling it and wrapping a coil around the outside that I would have a working speaker. He though about it for a minute and bet me that no sound would come out because there would be no linear travel possible by the voice coil. I still have that beer can and it still measures +/-3dB from 300 to 10K.
Decware is a trademark of High Fidelity Engineering
Co. |
