Chapter 3: Amplifiers

The most important thing to know about power level when selecting an amplifier is your speaker sensitivity and how loud you listen. Many people buy amplifiers that have far more power than what is needed. This is not necessarily a good thing. I recommend spending $50 and buy a good sound pressure level meter. They have been confirmed to be very accurate, certainly accurate enough for this job and certainly better than guessing. As a second recommendation, download an SPL meter app from your smart phone. While we may not know how accurate it will be, as it depends on the app, at least it will be better than guessing. I notice when people guess their SPL, they usually guess at least 10 or 20 dB higher than it is and that would unfortunately lead to them using an amplifier that is much more powerful than needed. Therefore, I cannot stress how important it is to know within a reasonable degree of accuracy your sound pressure level listening.

If you listen at the same level that your speaker sensitivity is rated at, that means you are using 1 watt. For example, say you measure your sound pressure level at your listening position, or at 1 meter from the speaker, you will find it is probably about the same unless you listen at a very great distance. If you measure 90 dB, and your speaker is rated at 90 dB for 1 watt at 1 meter that means you are using 1 watt. Of course, you want some headroom so if you listen at 1 watt average, a 10, 20 or 30-watt amplifier is certainly adequate to the task. There is no need to get 100, 300, or 400 watts if you are listening at just a few watts. Larger amplifiers do have problems playing lower power. Nelson Pass (First Watt) vehemently believes that in many cases 1 watt is perfectly adequate. It turns out when you get a sound level meter and start looking at your listening level, you will find that high levels are very unusual. If you are a person who listens at 100 dB, then you may want to consider a larger amplifier or more sensitive speaker. There are very few applications where over 100 watts is necessary. Think carefully and do some measuring before you go out and buy a 500-watt amplifier.

The next thing you need to know when selecting an amplifier is the impedance curve of your speaker. This is often published, or you could find the specification for the minimum and maximum impedance. Some speakers, magnetic planar speakers for instance, have very constant impedance and we do not need to worry about the peaks and dips. There are other speakers where there are very high impedance peaks, which usually happen at the resonant frequency of the speaker at the low end around 40 Hz to 50 Hz. There are often dips in the middle somewhere around 1k Hz to 2k Hz and then the impedance often rises again as you go up in frequency. Some people design speakers with an impedance that varies a great deal. As an example, 40 or 50 ohms at low frequencies, which is not unusual, and then dipping down to 3 or 4 ohms at some other place in the curve. So, it is also important to understand the damping factor of the amplifier. To understand damping we need to also understand the application of feedback. Feedback in general has been given a very bad and undeserved reputation. There are people making zero-feedback amplifiers because customers out there are afraid of feedback and so they buy a zero-feedback amplifier. They do not know what feedback is because nobody has really told them. So, I am going to tell you.

Feedback is a simple system where we compare the input signal to a small portion of the output signal and correct it. Basically, the input comes into the amplifier and the feedback is there immediately. Some people use the claim that feedback comes after the fact so they cannot correct the music because it comes after the music. However, it comes simultaneously with the music. When you turn your car lights on there is not appreciable delay between the turning on of the lights and you then seeing the lights. It travels at the speed of light and so do electronics. Music moves so much slower than the speed of the feedback that we do not need to worry about this. It is true there can be delays in feedback. If those become a problem then you have bad feedback and then we have an amplifier that is too slow to correct itself in time, and that would be a very bad amplifier. In solid-state amplifiers, feedback is almost mandatory because solid-state devices are not that linear. We must come up with some way of linearizing them. Zero-feedback amplifiers do not have incredibly high damping. They do not have incredibly low distortion. They can have wide bandwidth, but designing for bandwidth is not a problem. Adding a few dB of feedback makes a small change, but if you add 6 dB of feedback, you do three things. You lower distortion by half, you double the damping factor, and you increase bandwidth by a full octave on both ends and that is quite a desirable thing to do.

I read recently about someone making zero-feedback amplifiers that feels that is the only way to make amplifiers. This designer also said that they added 2 dB of feedback to their amplifier, and it just ruined it. Frankly, I think that is not true. I think that is just an advertising lie and a marketing tool as 2 dB of feedback cannot make any difference in any direction, and in my opinion this statement discredits the designer. Another example of feedback lies is someone who made a very expensive $80,000 vacuum tube amplifier and put a switch on it that would add either 2 dB of feedback or let the amplifier operate with zero-feedback merely to demonstrate that switching it to 2 dB of feedback would ruin the sound. I think that is the most ridiculous thing I have ever heard of. Why would you put a switch on an $80,000 amplifier just to prove a silly point like that?

There are situations where feedback in an amplifier can cause a problem due to poor design. The more feedback you put into an amplifier and the less skill the designer has you are likely to run into these problems. I believe this is one of the reasons that some amplifiers and speaker combinations sound horrible. One example of this is what we refer to as a birdie. It is called a birdie because if you look at the sound wave on an oscilloscope it is a continuous fine line. A birdie is a bleep on the line, and it looks like a bird sitting on wire. This is an engineering term that is well understood by engineers who understand feedback. What is going on with the birdie is that at that point in the sine wave, and that could be a particular point of the musical wave as well, the amplifier literally bursts into oscillation. The oscillation though is of such a high frequency that you cannot hear the pitch. It is perhaps a megahertz, but that oscillation creates bad distortion that sounds a lot like clipping but is worse than clipping. I have found it is most likely excited by a woofer. When you get near the resonant point of the woofer, the phase angle is such that it is a very sharp-phase angle and this along with the back EMF of the woofer can cause these birdies.

Many times, I have designed an amplifier, and it has passed all my tests for different capacitive loads, inductive loads, and every other kind of load I can throw at it. Yet I put it on the woofer and crank it up to some reasonable levels and sweep the woofer over different frequencies, and while it only happens at certain frequencies, here come the birdies and it is quite audible. We are not talking about a subtle thing. In music, it is not as easy to hear because when you are doing this with the sine wave you have this pure sign wave and then when the birdie comes in you get quite a few harmonics on top of the sine wave. It literally has a buzzing sound. This is an instance where poorly applied feedback causes a problem, and it is evident the designer of the amplifier never conducted the right tests to see or hear the problem.

Another problem with feedback when it is poorly done is that some designers feel that if you put the standard capacitive load on an amplifier, which is basically to add a 2-microfarad capacitor in parallel with the 8-ohm load and you then put a square wave onto the amplifier of 1 or 2k Hz, it is going to ring. You are going to see a ringing on top of the square wave, and it may burst into oscillation. They feel if they apply feedback, and they fix that problem then with any other capacitive load the amplifier will be just fine. What I found is if you put very small capacitive loads on an amplifier, such as a 10th of a microfarad or 100th of a microfarad, that same amplifier that was stable at 2 microfarads will not be stable at the lower capacitance loads. I expect that many designers do not bother to test their amplifiers at the lower capacitance, and they just assume if it is going to be okay at 2k Hz it will be good at anything less than that.

The problem with this is that some speaker cables now have enough shunt capacitance because they are coaxial that they can present a 10th or 100th of a microfarad load. I have seen amplifiers catch fire when connected to certain high capacitance speaker cables because the amplifier went into oscillation. When an amplifier goes into oscillation and becomes unstable, the input power goes through the roof. The output transistors must handle all that power internally and everything gets hot and then generally burns out, sometimes quickly. Many amplifier designers claim their amplifiers are stable into any load. However, I have had amplifiers on my bench where I put a 10th of a microfarad on them, and they are ready to blow up if I leave it on there for more than a few seconds. A designer saying that their amplifier is unconditionally stable might make for a nice  advertising statement, but it is not necessarily the truth.

My closing comment about feedback is that when it is a problem it is because the amplifier was designed incorrectly. You might want to know that there are many amplifiers designed by people who are not skilled in engineering and feedback is very often done on a cut-and-dried basis, which means they build an amplifier, they add the feedback, they see how the amplifier misbehaves and then they start adding parts, generally small capacitors here and there, to try to fix the problems. The question becomes did they find all the problems and did they fix them properly.

We should now talk a bit more about damping factor. We must assume that most speakers were designed with high damping solid-state amplifiers, because that is what most speaker designers are going to use. Once you have an amplifier with a damping factor over about 20, it does not matter if that damping factor goes up to several hundred. A damping factor of 20 means that you have a very low output impedance somewhere around 0.05 ohms. In fact, in my world I recommend that we do not talk about damping factor, and we just simply talk about the output impedance of an amplifier. You have probably read that damping factor is simply the ratio of the speaker impedance to the output impedance of the amplifier itself. Let us say you have an 8-ohm speaker, and you have an output impedance of 1 ohm. That would give you a damping factor of 8. If that output impedance was a 10th of an ohm, which is more likely with a transistor amplifier, that damping factor would now be 80. If it was 100th of an ohm, that damping factor would now be 800.

You can then sit and say, "I have an amplifier with 100th of an ohm output impedance," but now you add your speaker cable, which may be a few hundredths of an ohm, and the cable may cut the damping factor in half. Again, we do not really need damping factor much above 20. Feedback can be a very useful technique in amplifier design. As mentioned previously it reduces distortion, widens frequency response, and increases damping factor, or rather reduces the output impedance. One way to think about low-output impedance is to think about a small battery versus your car battery. Your car battery has very low output impedance and that is why it is able to start your car. When you turn the key and engage the starter, you are drawing an enormous amount of current. We need the battery voltage to hold up while drawing that large current. What happens to most car batteries as they age is not that they lose voltage it is that they lose the ability to supply current.

To state this another way. As the internal resistance goes up, and it goes up because the plates in the battery have decayed, parts have plaqued off and dropped to the bottom of the sediment chambers and you no longer have the surface area. The output impedance of a battery is totally dependent on surface area. The same thing is true of capacitors. A large capacitor has more surface area, more surface area means more current and lower output impedance. Almost always output impedance and current go together. If you have something that can supply a lot of current, it has a lower output impedance.

Now let us look at classes of amplifiers and their configurations. There are two ways to configure an output stage of a tube amplifier. There is single-ended and there is push-pull. Single-ended will have one tube doing all the work. There may be multiple tubes, but they are connected in parallel, so they function as just one larger tube. Let us just consider a single-ended amplifier to have one output tube such as a 2A3, 300B, 245 or, 211. These are the popular single-ended output tubes. Generally, single-ended amplifiers, due to their low efficiency, put out very little power and generate quite a bit of 2nd harmonic distortion. Larger tubes are simply used to get more power. You can usually get about 25% of the dissipation factor of the tube as power output, so a 300B usually cannot give you more than about 10 watts and most of them are 5 to 7-watts.

Single-ended amplifiers work by having the idle current in the output tube at 1/2 of peak value. If you have a tube with a peak value of 200 mA, then ideally you want to run it at 100 mA. While the tube is going to peak at 200 mA, it will also go down to 0 and that is the reason it must be run in the middle of the range. It must be able to go up an equal amount from where it can go down. If it was already biased at a very low level such as 0, it would be able to go up in current but there would not be any place for it to go down. The way this was solved is to make an amplifier push-pull. In the push-pull amplifier we have one tube that is going in the positive direction and another tube going in the negative direction. Push-pull amplifiers always have pairs of tubes. They may have multiple pairs of tubes, but you must have at least one pair of tubes to make a push-pull amplifier.

Then we get into the issue of the class of amplifier, which is where we have classifications such as A, AB, and B. A single-ended amplifier must always be Class-A, which means you are running a very high idle current. Some people like to call this bias, but it is more correct to call it idle current. Bias is the negative grid voltage that you put on a tube to get the idle current that you want. For instance, if we talk about a 300B amplifier, the grid will be negative 60 volts to get about 60 to 80 mA of plate current, which is a very typical operating point for a 300B. A 2A3 operates at similar numbers, but a little bit lower, and something like an 845 is going to have several hundred volts of negative grid bias and probably run 100 to 150 mA of plate current. That is why it can put out more power because it has a higher plate supply voltage and a higher plate current.

Push-pull amplifiers can be biased at different levels. Again, the bias is the negative voltage we place on the grid to get a certain plate current. You can make a Class-A amplifier that is push-pull and that merely means that the signal current never exceeds the idle current. This limits the power of the amplifier because what is going to happen is you must run the idle current high enough, but running the idle current high heats the tubes and the tubes can only take so much heating. Basically, a pair of KT-88 tubes running pure Class-A might be able to do about 20 or 30 watts, possibly 40 but that is pushing it.

Now an amplifier designer can take a pair of KT-88 tubes and get as much as 100 watts out of them, but it will no longer be Class-A. To get the 100 watts, you need to lower the idle current from say 100 mA down to say 30, 40, or 50 mA. These are the numbers that you are familiar with when you adjust bias, and you measure for millivolts across a resistor. Every millivolt represents 1 mA of current, so you are really measuring current when you are adjusting bias. People often say they are running 30 mV of bias on a tube, but that is an inaccurate statement. What is really happening is they are running 30 mA of idle current. The bias running on the tube is unknown. You have adjusted the grid to some unknown voltage to produce 30 mA of idle current.

Since the amplifier we are discussing can no longer be called Class-A, we instead refer to it as Class-AB1. What is really happening is that at low signal levels, and possibly for all the signal levels you are listening to, except for transients, you are running in the Class-A region. Then as the music gets louder you move into the B region and that is why they are called AB amplifiers, because at some signal levels you are in the A region and at higher signal levels you are into the B region. The Class-B amplifier is a rather misunderstood amplifier. It is one in which there is no idle current. Again, people like to say there is no bias, but the truth is there is no idle current. It is very hard to say what no idle current is because there is going to be some various small amount of current. Let us say when you get below 1 or 2 mA, we refer to that as no idle current. The problem with this is that when you switch from the positive-going tube to the negative-going tube there can be a little dead zone, and this is what is called crossover distortion. This was also an early problem in transistor amplifiers for the very same reason in that they just did not run enough idle current. You bring up the idle current to the point that you minimize this transition region so that it hardly exists any longer and now you have a Class-AB amplifier.

There is one more thing to talk about and that is the difference between a Class-AB1 and Class-AB2 amplifier. Class-AB1 and Class-AB2 just refer to tube amplifiers. Transistor amplifiers are all Class-AB1. A tube amplifier can become an AB2 if the grid is able to be driven into the positive region. In a regular Class-AB1 amplifier, the grid never goes above 0. It is always negative value. At peak signal it comes up to 0 but never exceeds 0. If you exceed 0 on the grid and go positive, you could get extra current out of the tube and there are some amplifiers that are made to go into this region. The beauty of this is that when your speaker impedance drops and you need extra current to drive that lower impedance, the amplifier can provide it. That is the main advantage with a Class-AB2 amplifier and there is no disadvantage in having that. It is a little harder to design and there are companies who design them, but most companies just do not do it. It is not in their way of thinking.

Class-B amplifiers never existed in the original vocabulary of amplifiers, which basically were all designed using today's tubes. We just adopted the tube designations to modern transistor amplifiers. We do things like this often in engineering. A Class-D amplifier in its modern definition could also be called a switching amplifier. A switching amplifier does a very different thing from a linear amplifier. It is not digital, and I think it is a mistake to call a Class-D amplifier digital although we have “D” for digital and “D” for Class-D. I do not know why “D” was chosen except that it was the letter following “C” and as far as I know, we do not yet have a Class-E amplifier.

A switching amplifier is generally done with transistors although you could design one with tubes. in a switching amplifier the output transistors are being switched either completely on or completely off. Imagine we have this device like a light switch, and we can switch it on or off and nothing in between. The beauty of this is that transistors, when they are either fully on or fully off, generate virtually no heat. This is because when they are fully on, although there may be a lot of current, there is extremely small voltage, and the dissipation is always voltage times current. So, we can have a lot of current with very little voltage and no heat. When the transistor is off there is a lot of voltage but there is no current, so again no heat.

The reason that conventional transistor amplifiers get hot is because there is always voltage and current present across the transistor. There is the bias idle current and then there is the music current. Of course, if you have an amplifier that runs cool at idle, as you exercise it and make it do some work it will get hot. If you have a Class-A amplifier, it tends to run hot all the time because the signal does not make much difference. In fact, the Class-A amplifier runs cooler as you play it at full power because now while the input power going in is the same, you also have output going to the speaker. The output going to the speaker subtracts heat from the Class-A idle current.

In a Class-D amplifier, you nearly always have a high gain op-amp, and a great deal of feedback. These are almost impossible to avoid. The way a Class-D amplifier works is that the comparator, the switching of the output transistors on and off, is looking at the average voltage going to the speaker and comparing that average voltage or a fraction of it to the input voltage and again simply matching them up as feedback always does. In reality, if feedback were a bad thing, then Class-D amplifiers would be the very worst of them. Indeed, a Class-D amplifier is in a way proving that feedback is a good thing and not a problem. What happens at this point though is if the amplifier has a sound at all, we will have to say that the sound of the amplifier is now determined by the quality of the input op-amp. If the input op-amp, or the components associated around the input op-amp have any sort of phase distortions or voltage level distortions, those will appear at the output.

As I said earlier, to make a stable amplifier, we must have something that is very wide bandwidth and with very little phase shift so that we can literally wrap some feedback around it and have the feedback signal which is going to be compared to the input arriving simultaneously. Some people like to think there is propagation delay in the amplifier. I encourage people who think about things in the time domain to suspend that type of thinking for a moment, because the time domain can get you into lot of trouble in your thinking. What happens is not the fact that it is delayed but the fact that there will be a phase shift at some high frequency. It is a fact that the frequency of the phase shifts as much as 90 degrees, and may even shift to 180 degrees. At this point the amplifier now becomes an oscillator and that is the key problem with feedback. The good amplifier designer will make sure that the amplifier always stays at 90 degrees and never goes to 180 degrees.

When this happens it causes an amplifier that is stable on the bench to oscillate when connected to a speaker where the speaker load, or the cable load, can add an additional phase shift that the designer did not consider. As was mentioned earlier, many amplifier designers only consider the resistive load. That is why most amplifiers with very high feedback, and almost all transistor amplifiers, have a very small output choke to decouple the output of the amplifier from the load at very, very high frequencies. Some people fear this little output choke that is a simple little coil of 18-gauge wire of maybe about 10 turns. Again, there are designers who decided that this is an evil little part, and they are going to tell you this is an evil part, and they are going to try to sell you their amplifier because it does not have this evil part. I want you to know there is nothing evil about this little, tiny coil of wire that people are working so hard to eliminate. Frankly, whenever I open an amplifier and I see that nice little coil of wire in there at the output terminal of the speaker binding post I go, "That is a very smart designer."

In addition, a smart designer will put a parallel capacitor with a resistor in series that is called a Zobel network. The Zobel network is merely there to present a load at high frequencies which again helps stabilize the amplifier. The designer of the amplifier may not know at the other end of your speaker cable what that speaker load looks like and at very high frequencies most speakers look like an open circuit. The Zobel network has about a 10th of a microfarad capacitor and a 10-ohm resistor, and this presents a 10-ohm load at very high frequencies, such as above 100k Hz. One of the problems of the Zobel network that is interesting is that the resistor is usually only a couple of watts. If you test one of these amplifiers at full power and say at 50k Hz because you are having some fun with it on the bench, the resistor often starts to smoke and catches fire. This has happened to me with a particular amplifier, and I called the manufacturer, and we had a nice little chat about that. One of the problems with testing amplifiers is you must be a little bit careful of testing them at full power at very high frequencies.

In the 1950s, a very confident engineer named Julius Futterman, who worked for a test equipment company, designed an audio amplifier without an output transformer and this became known as the OTL or output transformerless amplifier. His circuit was very well written up in the Audio Engineering Society Journal and then later in one of the electronics magazines as an actual construction project and many people built these amplifiers. The early ones used a 12P4 triode and it was a very interesting circuit. You needed a whole bunch of these tubes in parallel because while they were the right kind of tube, they were rather small in their ability to put out much power. We also must realize that the speaker loads at that time were 16 ohms and OTL amplifiers tend to like higher impedance loads. The reason for this is that a vacuum tube unlike the transistor is more of a current limited device. For any given vacuum tube there is only a certain amount of current it can conduct.

The modern tubes being used for OTLs, such as the 6LF6 can conduct about 1 to 1 1/2 amps of current per tube. Compare that to a transistor that can conduct 10, 20, 50, or 100 amps and we clearly see that the tube has much less current capability. With an OTL amplifier we can put several of these tubes in parallel and the currents will add up and after you get enough current you can generate close to 100 watts at 8 ohms. The problem is if these OTL amplifiers are put onto low impedance speakers or speakers that have a severe impedance dip, they are going to overheat the tubes. If the amplifier is played loud, the tubes can get red hot, and the tube life will be significantly shortened. There is a common thought that OTL amplifiers and electrostatic speakers are a match made in heaven, but this is not entirely true. As audio myths go, this started with the Futterman OTL amplifier and the KLH-9 speaker. However, the KLH-9 speaker is a 16-ohm speaker with very constant impedance that does not dip much below 12 ohms whereas other electrostatic speakers can go as low as 1 ohm or 2 ohms making an OTL amplifier unsuitable for that speaker. Even a QUAD ESL dips down to a few ohms, but at an extremely high frequency, so an OTL amplifier is not a bad choice. However, an OTL amplifier can put out very high voltage and speakers like the QUAD ESL will arc above 35 volts. So, you are getting into the area where the speaker may arc and damage the panels.

Julius Futterman being a very clever engineer realized a few things about his amplifier. He realized that because there was no output transformer, he could put an enormous amount of feedback on the amplifier. In the article he spoke about the fact that if you put in enough feedback to make the amplifier have unity gain, meaning a gain of 1, you will have an incredibly high damping factor of several hundred or almost a thousand. When damping factors get that high it gets difficult to measure, and after damping has passed about 20 or 30 the extra damping does not matter. Futterman was also interested in high-level feedback because it allowed him to simplify the power supply and reduce any supply ripple to very, very low values.

The Futterman OTL circuit is very different than the Circlotron OTL circuit, popularized by a company called Atma-Sphere. The Circlotron circuit originally came from Electro-Voice and was a bridged amplifier. It is a bit more complicated circuit that may or may not have feedback. Most people who make Circlotron amplifiers do not employ any feedback and the damping factors of those amplifiers can be quite low. In fact, many of them are as low as 1, which means you have an 8-ohm output amplifier with an 8-ohm internal impedance and virtually no damping. When you connect an amplifier with no damping to any speaker that has a rising impedance, you will have rising response at that frequency. I measured a Circlotron-type amplifier connected to a Quad ESL 63 speaker and found at 50 Hz the bass was up 8 dB which is quite noticeable.

At this point we should discuss the transistor amplifier. A virtue of transistor amplifiers is that they can supply a great deal of current. Of course, to supply this current a large transformer is needed but even more important is you need large filter capacitors because the capacitors are what are going to supply this instantaneous current that you need. There is no reason a good transistor amplifier with modern transistors, even at a low power level such as 10, 20, or 30-watts, should not supply at least 20 or 30 amps of current. If we go back to older amplifiers of the 1970’s and 1980’s, this was not true because transistors had not yet been developed for very high currents, but now we have very high current transistors. MOSFETs are extremely high current devices and very inexpensive. You can buy a 100-amp MOSFET for a couple of dollars.

A 100-amp bipolar transistor is much more expensive and 30 years ago that transistor was not even available, so a great number of transistors had been put in parallel to make such an amplifier. One example is the Mark Levinson ML-2, a low voltage but extremely high current amplifier. The ML-2 was designed to drive the HQD system with stacked QUAD ESL speakers hooked up in parallel. These speakers presented a load requiring a fair amount of current. Levinson also wanted to design an amplifier that would not use any current limiters. Current limiters are one of the biggest problems of transistor amplifiers. They are there for the sole purpose of protecting delicate transistors. Even though a transistor can supply a great deal of current, if it is supplying this current at the same time there is a great deal of voltage across it, you run into something called second breakdown. Some people call it secondary breakdown, but that is not correct. It is just called second breakdown and I believe it is called this because it is a second breakdown mode.

Transistors have several modes of breakdown the most popular one being voltage. If you over voltage a transistor, you will breakdown the junction. This happens in a fraction of a millisecond and there is no recovery from it. One of the virtues of vacuum tubes is that they can handle extremely high overcurrent pulses and have virtually no damage, certainly not shorting out and being made unusable. In fact, RCA even mentioned that one of their horizontal output tubes could go 400 percent over current and dissipation with no damage whatsoever. Frankly, on good output tubes you could go 10 times overcurrent very briefly and not damage the tube at all, but you cannot do this with the transistor. If you exceed any of the ratings on the transistor either voltage, current, or second breakdown, that transistor is done.

Another point of interest regarding damaged transistors is that if one transistor fails in an amplifier it often takes another transistor with it. For example, if one output transistor shorts out because it is overstressed due to a current peak or short circuit in the speaker leads, that transistor now tries to pull the output pin of the amplifier, for example, the plus output terminal. The overstressed transistor tries to pull the output pin up while another transistor is going to try pulling it down against the short. It will then overcurrent that transistor so now you have a pair of shortage transistors. Hopefully, it blows the fuse at this point and does not do any further damage. Quite often in some transistor amplifiers, there is this whole cascading effect where the output transistors fail causing driver transistors to fail as well. This is why the repair of solid-state amplifiers can be a rather expensive thing. With a tube amplifier, if one tube fails it does not harm any of the other tubes. You simply replace the tube that has gone bad, and you are back up and running again. This is one of the things that I happen to really like about vacuum tube amplifiers. Also, vacuum tube amplifier output devices are very tolerant of short-term high current overloads, so they do not need any protection circuits.

Protection circuits have another problem besides protecting shorted speaker leads or impedance currents that go too low or someone hooking up too many speakers in parallel which again can draw out too much current. In some cases, the limiters can be tricked into coming in. Now some amplifiers, especially Japanese receivers, have done the protection mode with a relay. There is a circuit looking at the voltage and the current and deciding if there is too much current at a particular voltage. It opens the relay, disconnects the speaker, and protects the amplifier in that way. In theory, this is a better way of doing it because when the relay opens the amplifier shuts off. These types of limiters are not used so much in high-end amplifiers because we do not like the relay contacts. Over time the contacts oxidize and then the oxidized contacts create distortion. Again, in high-end amplifiers we want the most direct path possible from the output transistors to the output terminals of the amplifier. We can be pretty sure that most high-powered solid-state amplifiers are going to have limiters because they work at higher voltages and higher voltage transistors are more difficult to protect. Also consider that when you buy a solid-state amplifier of more than 100 watts you are getting into “race car” technology where things are going to be more delicate, fussy, and likely to break. If you do not need the extra power, you are better off not having an amplifier of that size.

Load line limiting in transistor amplifiers is probably one of the greatest causes for what we might call a mismatch between an amplifier and a particular speaker. If we take a transistor amplifier that has sensitive load line limiting and connect it to a speaker with very low impedance or a very large phase angle, this would tend to trip the load line limiters at higher levels. As you are listening to the music, what is going to happen is when the load line limiter kicks in it literally causes clipping and it is instantaneous. It does not sit there and tell you it is doing it. It just does it, and it sounds about the same as clipping or slightly worse. It does not shut the amplifier down, but rather stops the output. It interrupts the output briefly until the speaker load or the music content material gets down to a low-enough current that the amplifier can handle it.

The worst kind of load as far as a transistor amplifier is concerned is the electrostatic speaker. The electrostatic speaker is typically a straight capacitive load, which means it has a 90-degree phase angle. What this means is that when the amplifier is putting out the most current, it also has the most voltage across the transistors because the voltage across the speaker is 0. From the transistor's point of view, which is what the load line limiters are looking at, if there is no voltage across the speaker, then all the volts must be across the transistor. Now the load is asking the transistor to deliver a great deal of current while all the volts are across it, and it is not going to do it. It is as simple as that. With the resistive load the output voltage and output current always go hand in hand. When the output voltage is the largest and the voltage across the transistor is least, the transistor is quite happy. The limiter is looking at the voltage across the transistor, the voltage across the speaker, and the current instantaneously and continuously. It is doing this with a very simple circuit. There is not a computer or microprocessor in there. There is literally one very small transistor with a couple of resistors sitting there continuously looking at the voltage and current relationship and protecting the output transistor(s).

Now let us consider the term “bridged amplifier”. This is a confusing term to a lot of people. When you bridge a transistor amplifier, it is quite different from bridging a tube amplifier. In bridging a solid-state stereo amplifier into mono, what you are doing is simply inverting the phase of one of the channels and connecting the speaker between the hot terminals of the two channels. What happens is that when one channel is going up or positive, the other channel is going down or negative. The speaker sees twice the voltage that it would have seen if the amplifier were not bridged. What this does is produces four times the power. Power is voltage squared divided by resistance. However, you do not always get four times the power because now you are loading the amplifier with half the load impedance of the speaker. As far as the amplifier is concerned, that 8-ohm speaker now looks like a 4-ohm speaker. If you were to imagine the speaker having a center point, that center point would be at ground potential and not moving. Literally, you are driving one lead of the speaker positive while the other negative and doubling the voltage.

If you want to know how much power any amplifier puts out in bridged mode, you look at the power it puts out into half the load impedance that you are using. For instance, if you are using an 8-ohm speaker, you look at what the 4-ohm power would be (the total power of the two channels running) and that will be your 8-ohm power into an 8-ohm speaker. Again, the amplifier thinks it is seeing a 4-ohm load, which is why most bridged amplifiers will not tolerate a 4-ohm speaker because if you put a 4-ohm speaker on the bridged amplifier it thinks it is a 2-ohm load. If you are going to bridge an amplifier and run a 4-ohm speaker, you better make sure that it can handle 2 ohms in stereo mode. This is a very important thing to consider.

People talk about bridging a tube amplifier and that is a term that should never be used, because when you take a tube amplifier and you turn it into mono you are not driving the two channels out of phase, like you do in a transistor amplifier. You are driving the two channels in phase but in parallel. The proper term would be paralleling the channels, not bridging them. Now a whole different set of things is happening. If you connect the 8-ohm taps together and you connect the ground taps together, which may already be connected, you now have a 4-ohm amplifier. If you still use an 8-ohm speaker, you are literally not going to get anymore voltage, but you will have more current available, but your speaker will not necessarily draw more current so you may not get any excess power at all. The power output of a mono tube stereo amplifier may be about the same as its stereo power output. However, if you have a 16-ohm tap on the amplifier and you connect those in parallel, then connect them to an 8-ohm speaker, you will get the total power of the two channels combined into the one speaker. The rule of thumb for paralleling a tube amplifier is always connect the taps that are twice the impedance of your speaker to get the full power. If you do not need the full power, you do benefit by connecting the two 8-ohm taps together and connecting them to an 8-ohm speaker because that doubles the damping and that makes a sonic difference, and it also doubles the current which helps you if the speaker has an impedance dip. That is the advantage of paralleling a tube amplifier.

There is also another problem. If the two channels of the amplifier did not have extremely well-matched gain, and I would say it needs to be at least a 10th of a decibel in gain matching, the two channels will fight each other, and things could be worse in mono than they would be in stereo. In fact, the Dynaco Stereo 70 amplifier has a mono switch on the front, the reason for which we do not know, it could be for playing mono records, or it could be for another reason. If you switch a Stereo 70 into mono, you must also parallel the output terminals. Dynaco suggested in the manual that you put a small resistor in series with each output terminal before you tie them together to take care of the fact that the two channels may not be gain matched. However, because you have added resistance, there is now loads of damping factor to the point that you are probably back to where you started with the damping factor, so it is not such a great idea. There are some amplifiers, like the Music Reference RM-10 where the mono switch matches the gains of the two channels and will always provide a balanced matched output.

Amplifiers on the input end come in two flavors now, balanced and unbalanced, the latter which people are also calling single ended. I find this is a bit of a mistake as they are calling the RCA-type input, which is really an unbalanced input, single-ended. I think it is a little confusing because we also have amplifier topologies called single-ended and I think the term single-ended really belongs to an amplifier topology, not to a type of input. This is the audio community making up their own vocabulary which often is not the vocabulary that has been established by the engineering profession. The benefit of a balanced input, which of course is very popular in studio equipment, is that it rejects any noise that is injected into the cable. You can see in the studio this makes sense because there are often very long cables run from microphones to the mixing panel at low level so that injected noise could certainly be a problem. In home systems, our cables are generally short enough and well shielded that we do not really need to use balanced inputs, but balanced input provides one other advantage and that is the breaking of ground loops.

If you look at classic equipment, such as the early Marantz, McIntosh, and Fisher amplifiers you may recall that none of these have a ground pin on the power cord. Sometime around the 1980’s people started putting ground pins on power cords because we now have grounded outlets in houses. The ground pin on the power cord is merely done for safety reasons. The safety factor is if internally the power transformer or something inside the amplifier touches the chassis of the amplifier or shorts the chassis of the amplifier, the chassis can become what we call “hot”. If you were to put one hand on that chassis and another hand on the cold-water pipe, you would receive a very dangerous shock. By grounding the amplifier, if the amplifier now fails internally, it blows your circuit breaker and then you will know there is something wrong. You can retain all this safety by merely grounding the power amp. In any system, it is only the power amp that should be grounded. Unfortunately, once we started grounding one thing, we started grounding everything and that is where the ground loop problem arose.

Ground loops exist because of a ground path created by the interconnecting RCA cables, but we also have a secondary ground path created by all the ground pins on the power cords. In this loop, you can get a circulating current and that is why it is called a ground loop. We now have a closed path of magnetic flux coming from everywhere induced into that loop and causing a very small voltage which is carried on the ground of the RCA cable. So far as the amplifier is concerned, that looks like a music signal. All an amplifier can do is look at the signal of its ground relative to the input and if there is signal on the ground it is just as good as signal on the input. With the balanced input we now avoid this problem and the way we avoid the problem is that the chassis may be tied together through all the power cords. We now have a plus and minus signal coming through the cable and the amplifier simply looks at the difference between that plus and minus and rejects anything that is common to them. That is called common mode rejection.

If you are concerned about hum problems due to cables, you want to make sure the amplifier that you buy has what is called a high common mode rejection ratio, expressed in decibels and that should be about 60 to 100 dB. It merely means if you have a 60 dB common mode rejection ratio you reject the noise on the cable ground to the level of 60 decibels, which is generally enough. If for some reason someone made an amplifier with a common mode rejection ratio of 0, then the balanced input would absolutely do you no good at all because it is not looking at the difference. It is looking at the ground also. If you are using balanced, it is appropriate to ground all the chassis and not have the chassis floating. If you let the chassis float that could exceed the common mode input range of the receiving amplifier and cause other problems. Generally, that does not happen because the XLR-type connector carries the ground through on the shield so the chassis will be tied together, and you can confirm this with an ohm meter. We do want to make sure that the chassis are all roughly the same potential to give you hum rejection.

If you are using separate mono amplifiers with RCA-type or balanced inputs and have hum issues the common thing to do is to float the ground on one of the power amplifiers. It is about the only thing you can do. Just because you plugged the two amplifiers into the same line or to the same power socket that does not mean you are going to get rid of all the hum. If you think about it and you look at the cabling on the floor and you say to yourself, "I have got a big circle of ground here," then indeed you can have a ground loop and you can have hum. This has always been the problem with using separate power amplifiers although it is nice to put the amplifier next to the speaker, but now you have the potential of the ground loop.