------------------------------ From: stevesaf@microsoft.com (Steve Safarik) Subject: Re: Tube Amplifier Improvement Date: Tue, 30 Nov 1993 21:43:14 GMT There are several things you might try to boost the low freq response, in rough order of importance: 1. If you're not using neg. feedback, use it. 2. Make sure that your B+ supply/filter capacitors are of low impedance. This means avoiding electrolytics (use poly. or oil/paper motor starting caps). 3. Make sure you have a decent output transformer. 4. If you have any caps in series with the signal (for stage coupling, or at the input to limit low freq response), try increasing their values. 5. Use large enough wire for the power supply. That's about all that comes to mind right now. Using poly filter caps should have a very big effect, but note that non-electrolytic caps are much larger, and you'll have to rearrange things to make them fit. ------------------------------ From: spaaejg@ucl.ac.uk (Eric Glover) Subject: Re: Tube Amplifier Improvement Date: Wed, 1 Dec 1993 16:21:49 GMT stevesaf@microsoft.com (Steve Safarik) writes: :There are several things you might try to boost the low freq response, :in rough order of importance: :1. If you're not using neg. feedback, use it. If you're not using NF you probably don't want to use it and may not have allowed for the reduction in sensitivity. :2. Make sure that your B+ supply/filter capacitors are of low impedance. This : means avoiding electrolytics (use poly. or oil/paper motor starting caps). Yes, but "motor run" not "motor start" caps are better for safety reasons. :3. Make sure you have a decent output transformer. Hear, hear! I speak from experience :-( :4. If you have any caps in series with the signal (for stage coupling, or : at the input to limit low freq response), try increasing their values. Cathode bypass caps will also roll off the bass. I try to use values that give < 10% of cathode resistor impedance at 20 Hz. :5. Use large enough wire for the power supply. :That's about all that comes to mind right now. Using poly filter caps should :have a very big effect, but note that non-electrolytic caps are much larger, :and you'll have to rearrange things to make them fit. Polypropylene caps also have the advantage that they are usually available in higher voltage ratings that electrolytics, so if you're having to put electrolytics in series the space disadvantages are not so great. Eric Glover ------------------------------ Date: Wed, 1 Dec 1993 12:31:00 +0000 From: "henry (h.) pasternack" Subject: Tube amp improvements / Woofer "speed". I'd like to submit a couple of comments/questions regarding recent topics on the bulletin board. 1) What happens when you put Farads of capacitance in your power supply. I'm not an expert on power supply design, but I know enough to throw some reasoned monkey wrenches into the discussion. Power supply filter capacitors serve two functions. The first is to attenuate 120Hz power supply noise (and its harmonics). They do this by shorting AC components on the supply rail to ground. It stands to reason that the lower the impedance of the power transformer and the rectifier diodes, the more capacitance you will need to achieve a decent level of atten- uation. Because of the inherently non-linear nature of diode switching, and the complexity of power transformer design, it is essentially impossible to predict analytically the performance of a capacitor- input filter. A long time ago, a fellow named something like "Schade" did some experiments and came up with a bunch of useful curves for filter design. In order to use his curves, you need to know the equivalent series impedance of your power transformer (which is not just the secondary winding resistance), the dynamic impedance of the rectifiers, the amount of capacitance, and the rectifier configuration (half-wave, full-wave, full-wave bridge, and so on). You also need to know your load impedance. The curves reveal several interesting things. As the capacitance is increased, a "knee" is reached where the gain in performance is maximum. Above that "knee", the benefit of adding more filter capacitance is not worth the cost. The reason for the knee, I believe, is as follows: When the reactance of the filter bank is large compared to the equivalent secondary impedance of the power transformer, the voltage across the filter during the charging part of the cycle increases rapidly. As the filter capacitance is made larger and larger, the charging time constant increases. At some point, the rate of charging is very long, the ripple is quite low, and the average output voltage is determined by the voltage division between the transformer secondary impedance and the load impedance. There is another effect to be concerned about. When a very low impedance transformer (i.e., one with a high current capacity) is used with exceptionally large filter capacitors, the duration of the charging cycle becomes very short. This means that the transformer secondary current consists of impulse-like spikes that have very high peak values. These spikes promote heating in the power transformer because their RMS-to-average ratio is high, and heating is proportional to RMS current. The peak diode current can be ten or a hundred times higher than the average current. Silicon diodes can handle high current peaks, but tube rectifiers cannot. This is why it is a really rotten idea to load a vacuum-rectifier amplifier with tons of filter capacitors, unless they are downstream of a filter choke, which tends to minimize the peak currents drawn from the rectifier. There has been a lot of fuss about fast-recovery diodes since the Greenberg follow-up article in "Stereophile". The reasoning is that these diodes sound better because they switch more cleanly. Given the brutal demands of modern "audiophile" power supply design, it's no surprise the rectifier switching characteristics can make a difference. Note that the peak current is, in the limit, the open-circuit secondary voltage minus the diode drop divided by the secondary series resistance plus the reflected primary resistance (which includes the wall socket impedance). This can be a hell of a lot of current. Aside from ripple rejection, the filter bank serves another purpose. Many people seem to assume that since the lower bandwidth of music is around 20Hz, the power supply capacitance should be selected with this frequency in mind. In fact, the spectrum of the power supply current variations can extend down into the infrasonic. This is because the current drawn from the supply is related to the power "envelope", which has little to do with the frequency spectrum of the music. Suppose, for instance, you are listening to pure tones at 1 kHz, and you begin twisting the volume knob at a 1 Hz rate. The major AC component of power supply current will be at 1 Hz. What is the reactance of your filter capacitors at 1 Hz? Pretty darned high. To some extent, large filter capacitors can "stiffen" the rail voltage when the power envelope is varying at a low frequency. (Editorial note: The tern "stiffening" capacitor seems to come from the car audio world. I hold a grudge against car audio- philes because the arms-dragging-on-the-ground crowd can't seem to relate unless they product is packaged with a hefty dose of macho sexuality. I just LOVE to stiffen my ten-inch Thruster.) Under these circumstances, the ability of the power supply to hold the rail steady has more to do with the regulation of the power transformer than the size of the filter bank. The same observation applies when listening to music with a heavy beat. Much of the perceived "slam" of a low bass note comes from high- frequency overtones. If you listen to the output of a subwoofer without the main speakers attached, the sound is pretty awful, really. It's hard to believe all that mush is really what it's cracked up to be. I think there's a lot more to getting good bass out of a power amplifier than adding extra filter caps. If the power supply in your amplifier is marginal to begin with, adding more capacitance can increase its workload and lead to overheating and reduced lifetime with no significant audible advantage. Power supply design is a complicated engineering problem. My recommendations for better bass? - Buy a reputable transformer, built with quality materials and having tight regulation, and operate it in the current range for which it is designed. - Use an ample, but not excessive, quantity of quality electro- lytic capacitors. - Consider Class A operation, which minimizes power supply current fluctuations, or, - Use regulated supplies, especially in the driver stages. 2) The use of poly bypass capacitors across electrolytics. It is often recommended that small polypropylene bypass caps be placed across electrolytic capacitors to improve their performance. The value often suggested is 0.01uF. I would like to point out that the reactance of a 0.01uF capacitor at 20kHz is around 800 Ohms. A decent, low impedance electrolytic has much lower impedance at this frequency. Adding a few tenths of a microfarad of capacitance to an electrolytic does not transform it into a "Supercap". Instead, you wind up with a complicated reactive network having potentially unpredictable characteristics. Plastic film capacitors have high Q and can tend to ring. It can be especially dangerous to place a tweaky capacitor on the output of an active regulator, because oscillation can be induced. If you look at the Curcio solid-state regulators, you will find a small resistor in series with the electrolytic cap at the output. This is to prevent phase shifts that lead to instability. Even passive power supplies can ring, however. What's nice about small film capacitors is that they are good RF bypasses. Monolythic ceramics are even better. Sprinkled carefully around a circuit, these small capacitors decouple the power supply lines and help to prevent radio frequency feedback that leads to RF instability. I'm not arguing that electrolytics are superb signal capacitors, but I do believe the audiophile press is overly simplistic about caps in general. I don't have the test gear right now to run experiments, but one day I hope to look carefully at this issue. In the mean time, I bet there are a lot of good articles out there, if someone cares to look. 3) Regarding woofer "speed". It's amazing, but not surprising, how few people seem to know how a loudspeaker driver really works. The mass of the cone is an integral, essential contributor to the proper operation of the system. The driving signal does not control the position of the cone, but rather, is related to the acceleration. Once in motion, the cone moves in an oscillatory trajectory that is determined as much by mechanical and acoustical resonance as it is by electromotive force. Over a range of frequencies and up to a certain SPL, the sound response is roughly linear with respect to the input signal. This is not so much because the cone is "controlled" by the driving signal as it is that the driver "cooperates" with the "hints" given to it. No doubt this is a crude analogy, but I am trying to drive a sharp wedge into the notion that the amplifier locks the cone into place as though by servo action. Speaker performance is largely a matter of inertia and damping. Think of the amplifier as a loud dog and the speaker cone as a flock of sheep. If the dog barks the right message, the sheep move where they are supposed to go. Try to move the sheep too fast, and all hell breaks loose. Watching the sheep dog, a child might think that the dog controls the precise position of each animal. On the contrary, any one of the sheep could scamper away at any time. It is because of sheep nature that they tend to do as they are told, and will move in a predictable way once set into motion. In groups of people, we call such behavior "inertia". The control an amplifier has over a loudspeaker cone is kind of like the control the sheep dog has over the flock. It's not absolute control by any means, but we still get to eat lamb chops when dinner time comes. Dick is right when he says this whole business of woofer "speed" is utter nonsense. The problem is that there are two definitions to the word, "speed". The first is an engineering definition, meaning the magnitude of velocity. Applied to loudspeaker design, there are well-accepted equations that predict cone speed in the presence of driving signals. The second is a subjective term, having something to do with perceived bass damping, lack of midrange coloration, and integration of bass with midrange and treble. The problem is that many audiphiles have inaccurate mental models of how loudspeakers operate. Loudspeakers are simple, but their behavior is counterintuitive. The conclusions audiophiles draw about the relationship between engineering "speed" and subjective "speed" are usually wrong in a factual sense. Maybe we need a new term to replace "speed". How about "Picard?" My subwoofer has a lot of "Picard". Make it so. -Henry