The Leach Amp FAQ

Concerning the Bias Current

From a correspondent on 2/7/03 - I have built your low TIM amp and am completely satisfied with its performance. Just one question. The quiescent current increases proportionally to the 8 ohm loaded output amplitude. Neither the frequency nor the waveform affects Iq, only the output amplitude on the 8 ohm load. Without the load, there is no Iq change. The following things have I checked: the VBE multiplier maintains its set DC voltage, the DC voltage difference between the power BJT's bases remains unhanged. Still Iq increases. For the measurements I used low pass filtering and ground-independent DMMs. I noticed this Iq change inside the amp across the 0R33 resistors and also outside, across a serial 0R22 resistor in the +58V supply rail. The amp anyway works perfectly, it has swept my old class-A JLH away, no other amp could do that so far. Please help me with some info on this Iq phenomenon.

With a signal on the load, you are measuring the bias current plus the load current. It sounds like your amp is doing what it is supposed to do. The peak load current can get quite large when driving a load.

Concerning Transistors and Sources

From a correspondent on 2/8/00 - I built your amp and it sounds great. I read in your FAQ-section, that one person asked you if he can use the European types MPSA06/56. I used the BC550/BC560 types and it sounds great. I also used BD243/BD244 instead of the MJE15030/MJE15031 and it works perfectly. But the Power of my amp is not 240 W, it is 310 W!!! Thank you for your perfect amp!!

From a correspondent on 1/31/00 - This will be my second Leach Amp, the first is fantastic! The first time around I made my own boards, no problems but once was enough fun for me. This one will be for a sub and a center channel. For your information, Allied Electronics proved to be MUCH, MUCH cheaper than MCM Electronics for almost all the ecessary transistors. A very good company.

How does the amplifier sound?

From a correspondent on 6/27/00 - First Excuse me for my bad English! I am from Bulgaria. I am 31 years old. Audio is my hobby all my life, especially constructing amplifiers. I have QUAD 303, 405MK1, PASS Labs ZEN, Douglas Self Class B amplifier all built by myself. Before 3 months I started building your Low TIM amplifier, and now the work is done. I can say one: this amplifier is the real GREAT sounding amplifier I ever listen. I make comparison with all other types I have, but Low TIM is the best one. I use good transformer which gives +/- 62 volts after rectifying and filtering. I use original components (it is not so easy to find original components here in Bulgaria, but I find it). And again: EXCUSE me for my bad English!

From a correspondent on 12/15/99 - I came across your web site a while ago and noticed that your Low TIM amplifier is still in production. I built two in about 1976 (immediately after the article in Audio ) have been using them ever since. I must admit it was the first time that I heard a very significant difference in sound quality by changing amplifying components.

From a correspondent on 12/8/99 - Hello! Now my amp works like a dream!! Thanks again for a great design. :-))

From a correspondent on 11/29/99 - Thank you.. the amp sounds truely wonderful. It sounds so good, I'll have to go through ALL my CD collection again.

From a former student on 11/2/99 - I was a student of yours in 1978, and I built the amp and set of speakers, which have lasted 21 years now. I work with a tube amp audiophile who used to swear that no IC amp could touch his tubes. I used the Leach Amp to drive his Thiel speakers and now he wants to build one. Please tell us how to order sets of the circuit boards, as I am ready to build a second amp for myself, just for the fun of it. Thanks.

From a correspondent on 7/11/99 - I have completed, with great success, two 3-channel Low TIM v4.4 amplifiers. I wanted to let you know how pleased I am with the sound of these amps - I have Carver amplifiers made in 1990 and 1996 and the Low TIM sounds so much better! Thank-you for a most stellar design and your efforts in keeping the Low TIM web page so informative and current.

From a correspondent on 7/25/99 - "My husband has been in touch with you regarding parts to build me a pair of Leach amps.

He tells you true, I did have a pair of your amps many, many moons ago. I was so bummed when I separated from my first husband. I have to tell you losing them amps really hurt...

BUT, I ended up with an amp, the Bedini 100/100 classic that I really liked. It is strong as a mule with a sincere attempt at reproducing musical reality, teamed up with my Ohm/Walsh speakers (I am a stickler for certain details such as Thiel alignment) and my R.G. Dynamics pre-amp (don't laugh now, I had Mr. Grodinsky assemble this piece of equipment for me long after he stopped making them as I had lost the original with the first set of Leach amps. It was the best pairing for the Leach's and my Revox turntable that I came across). I had a satisfactory set up. THEN I stumbled upon your current plans online. I copied them and sent my husband to work with them to look them over on his break. When he came home he downloaded all the necessary info and off he went, getting everything together to build a pair (YEAH!)

The Bedini set a standard some years ago for it's ease of handling of any and all sorts of music. I worked at a high end stereo store that allowed me to set the Bedini against any and all possible competitors. There was no comparison available on a retail level. Then I ran across a pair of your amps. I must say that when I A-B'd amps long ago against an original pair of your amps I was amazed at how fast the Leach's were in comparison. They do not 'stagger', or drag their 'feet' at all. It is something I think that is missing from all the top end amps I have heard to date, speed. When listening to a pair of Leach amps I truly felt like I was listening to music, not the electronic reproduction of it. So, if I must speak of a comparison, though the Bedini is a great, no a GREAT amp I will gladly set it aside (sorry John B.) to be able to enjoy what music should sound like.

I just want to thank you personally. If I never again, heaven forbid, get to listen to your equipment I have still spent many, many hours enjoying music thanks to your design. It is nice to know that there is someone still around who must understands music from the listeners point of view, not just from an engineering standpoint, making such great equipment available to the real music lovers."

From a correspondent on 7/15/99 - "I'll e-mail a photo back when it's done. My wife is the audiophile of the bunch--she had a pair of one of your earlier revisions some years back, but lost them when she separated from her first husband. She's using a Bedini 100/100 we bought used from a friend in Cal City (IL) right now, but claims yours sounded much better and has been looking for plans to rebuild a set for some time. I'm building the unit for her as a Christmas (99) gift."

From a correspondent on 5/10/99 - "Executive Summary: Wow!!! Details: It's been a while since I started this project. I built both channels, but I had no output when I connected a load. I ended up tabling the project for a _long_ time. It turns out I had interchanged Q10 and Q11, the protection transistors, and yes, I had made the same mistake in both channels. Boy, do I feel stupid. The first thing I played was the overture from "The Magnificent Seven." Oh my. Your amp also does well on Bach organ works. I expect the neighbors are going to call the cops on me pretty soon, if I don't destroy my speakers first. I had been using a 14W class A amp before this one, and I thought it was great, but I think this is better. It has its own sound, but overall it's a neutral sound, and it handles difficult passages effortlessly. Now I'm curious _why_ it's better. My listening room is small, and I usually listen at low to moderate levels, but maybe the class A amp was clipping. Or maybe yours is better because it's full-discrete vs. an OPA2604 in the front end. Or BJT vs. MOSFET. Or higher rail voltages giving a better small-signal approximation. Further listening and experimentation are required! Thank you."

From a correspondent on 5/10/99 - The first thing I played was the overture from "The Magnificent Seven". Oh my. Your amp also does well on Bach organ works. I expect the neighbors are going to call the cops on me pretty soon, if I don't destroy my speakers first. I had been using a 14W class A amp before this one, and I thought it was great, but I think this is better. It has its own sound, but overall it's a neutral sound, and it handles difficult passages effortlessly. Now I'm curious _why_ it's better. My listening room is small, and I usually listen at low to moderate levels, but maybe the class A amp was clipping. Or maybe yours is better because it's full-discrete vs. an OPA2604 in the front end. Or BJT vs. MOSFET. Or higher rail voltages giving a better small-signal approximation. Further listening and experimentation are required!

From a correspondent on 5/4/99 - "I would like to express my appreciation and thank you for the information in the "Leach amplifier web site." I have built two 3-channel units and enjoyed every moment of going through the phases of the project."

From a correspondent on 3/11/99 - "The amp is about 19 years old now. The speaker fuses blow occasionally because of dynamic range expansion processors. I want to thank you for creating such a unique amplifier. I've heard Amzilla, which I wanted really bad, and I could not afford it at the time. I did get to compare your amp to a couple of Dynaco tube amps and I think it compared favorably. I love the amp. I still do. I learned to fix amplifiers by your article and building this project. Thanks again."

From a correspondent on 3/8/99 - "This is a superb design. If you remove one power supply rail, the speaker terminals stay at zero volts. I don't know of any other design that would pass this test! Thanks again professor."

From a correspondent on 9/11/99 - "The Leach Amp was successfully built early this year without any problem, and it turned out to be better than the one I bought. It is not only my comment, but my friend also said the same thing. He is a DIY speaker builder. Three months ago, he borrowed my Leach Amp to test his Scan-Speak speaker system, he came back with very high rating on the Amp. Now, he wanted me to build one for him, and I misplaced your mailing address for ordering the circuit board. Please send me your address again."

From a correspondent on 8/23/98 - "I got your circuit diagram for the Leach amp off the web and have been meaning to write for a long time just to say thank you. I built one and have been listening to the amp for about two months now (very sharp reproduction). I decided to build another two amps to hear my speakers bi-amped."

From a correspondent on 7/20/98 - "I finally finished my first amplifier. They worked perfectly on power up. No smoke. These are very high quality amplifiers, IMHO. Thanks for designing a terrific amplifier."

From a correspondent on 1/13/98 - "I built one of your Low TIM amplifiers back in '79 when the article first appeared in Audio magazine. I have had a great amplifier for all these years. I had a Crown DC300 I sold in favor of your design and my construction. Of course I could be a bit biased but, I'm sure your design sounds better.

There are two connections for central ground on the board. The one on the input side is on the same trace as the signal ground. What is the purpose of second central ground connection if you are isolating the input jacks from the chassis and have R51 (82ohm) between input ground and central ground? It seems as though you worked on isolating the two grounds from each other and then shorted them with the second central ground connection.

For audio frequency equipment, a central ground is used and all ground wires run in parallel to this point. This prevents what are called "longitudinal voltages" in the ground system from coupling between different stages in the circuit. With parallel ground wires, there is the possibility of what are called "ground loops" if the ground wires connect together at either end. Thus the input cable ground is isolated from the chassis because it is grounded through the circuit board ground. There are 2 ground tracess on the circuit board, one in which signal currents flow and one in which ripple currents from the electrolytic decoupling capacitors flow. Each of these ground traces has its own wire that connects to central ground. A single ground should not be used because the ripple currents could induce hum in the signal part of the circuit. The 2 grounds are connected together with a 82 ohm resistor which is large enough to prevent the current in one ground trace from flowing into the other ground trace, but small enough to keep both traces at approximately the same ac voltage at higher frequencies where the inductance of the ground wires causes the ground system impedance to increase. The theory of this is covered books on electromagnetic compatibility.

Can I mount the main heat sinks inside the chassis with the tops of the power transistors facing up?

If you put the heat sinks inside the chassis, they should be mounted vertically with adequate vent holes beneath and above them. If you mount them horizonatlly, you should install a fan to keep them cool. I much prefer mounting the heat sinks vertically on the back of the chassis.

Does the amplifier put out a dc voltage on the loudspeaker if one of the power supply fuses blow?

No. If a power supply fuse blows, the amp goes dead and the output voltage is zero.

I have compared your circuit diagram to your parts layout for the circuit board, and I think that I have found an error. The order of R30 and D5 is reversed when you compare the diagram and the layout, and similarly for R31 and D6.

These elements are in series. The order of series elements doesn't matter. A math analog is 2 + 3 = 3 + 2. It doesn't matter which comes first. I have redrawn the circuit diagram to reverse the order of the resistors and diodes.

I do not need the 120 W power and I am going to remove transistors Q20 and Q21. Should I change any other elements, e.g. R36, R37/R38, or R41/R42? What about connecting C19 and C20 to R37 and R38?

Removing 1/2 the output transistors will not reduce the output power. It reduces the maximum output current by a factor of 1/2. For an 8 ohm load, the power output would not change. For lower load impedances, the protection circuits would kick in at a lower output current than with all 4 output transistors. If you want to operate the amplifier with 1/2 the output transistors, Q18 and Q19 are the ones to omit. You can also omit R37, R38, R41, R42, R45, and R46. These changes will reduce the maximum output current by a factor of 1/2. If you wish to decrease the output power, you must reduce the rail voltages. I do not recommend reducing them to less than +50 V and -50 V.

I have a transformer which gives dc rail voltages of +66 V and -66 V after rectifying and filtering. Should I change any elements, e.g. R34, R35, or R36?

The +66 and -66 V rail voltages are too high for the amplifier. Do not use this transformer. Some of the transistors could break down.

May I substitute European type transistors for the ones you specify, e.g. BC546/556 instead of MPSA06/56?

I am sorry, but I do not have any experience with these transistors. If they have similar specifications, they may (or may not) be suitable. I have included specifications for all transistors on the amplifier page.

Why do you use the 2N3439/2N5415? These transistors are designated for switching applications and may have bad linearity.

My Motorola data book says that these are "one ampere complementary silicon high-voltage power transistors." To my knowledge, these are some of the most commonly used complementary transistors for the gain stages and pre-driver stages in power amplifiers. I am not aware of linearity problems with transistors specified for switching applications. Some of these are very low capacitance and low noise transistors. For example, the 2N4401/2N4403 complementary pair are specified as switching transistors. They are some of the lowest noise transistors available for linear audio applications.

I now have a 40 W amplifier that was published in about 1985 in "Wireless World" or some similar magazine and was called a "Low TIM Amplifier." It has a slew rate of 40 V/us. The author specified the use of transistors in the output stage with a minimum gain-bandwidth product of 5 MHz to avoid distortions produced by the charging and discharging of the base-to-emitter junction capacitance. The MJ15003/15004 have a gain-bandwidth product of 3 MHz. What about that?

The higher the current rating of a transistor, the larger the junction area and the larger the junction capacitance. The driver transistors in the T-circuit output stage have a very low output impedance which can charge and discharge the junction capacitance of the MJ15003 and MJ15004 with no problems. The speed of the amplifier is set by the Q12/Q13 stage and not the output stage.

Why didn't you use current source tail supplies for the diff amps?

I had intended to use current sources in the beginning. However, when I started laying out the circuit board, I found that they added a great deal of complexity to an already dense part of the board layout. Therefore, I opted for resistive supplies to simplify the layout. I never had any problems with the resistive supplies. The amp seemed to work perfectly with them. An additional bonus is that the matching of the tail currents in the two diff amps is not a function of the matching of transistor parameters.

The turn-on characteristics of the diff amps determine the turn-on characteristics of the amplifier. With resistive tail supplies, the diff amps turn on very gracefully at a rate set by the RC time constants in the tail supply circuits. This leads to a thump free turn on, eliminating the need for output relays to protect the speakers. If transistor current source tail supplies were used, the current source transistors tend to turn on abruptly as the power supply voltages come up, which can produce thumps in the loudspeaker. This problem is even worse if one tail supply transistor turns on before the other.

Can I use separate power supplies for the two channels?

Yes. This will require separate transformers, rectifiers, and filter capacitors. An alternate is to use one transformer with separate rectifiers and filter capacitors for the two channels. My laboratory amplifier uses this configuration.

Can I mount each power transistor on separate heat sinks?

Don't do this. The amplifier will not be thermally stable. The four bias diodes should see the average temperature of all four output transistors. This occurs only when all transistors and all bias diodes are on the same heat sink. There is an exception. You can mount the two NPN output transistors for one channel on one heat sink and the two PNP output transistors on another. Either put all four bias diodes on one of the heat sinks or put two bias diodes on each.

I once had a student who used 2 heat sinks per channel with two output transistors on each, a NPN and a PNP. He put the bias diodes on only one of the heat sinks. His amp was thermally unstable. The heat sinks without the diodes overheated. Putting both NPNs on one heat sink and both PNPs on the other solved the problem.

Can I put the output transistors for both channels on the same heat sink?

Yes, but I prefer separate heat sinks. Put the output transistors for the two channels on opposite ends of the heat sink channel. The bias diodes for each channel should be in the center of the output transistors for that channel.

Can I power more than two channels from a single power supply?

Yes. However, the maximum power output per channel with all channels driven simultaneously will be decreased. The power output with only one channel driven will not be affected.

What is the difference between a toroidal core transformer and an E-I core transformer?

For a given power rating (or VA rating), the toroid transformer will be lighter. However, I have found that the voltage output of a toroid seems to drop more under load than that of an E-I core transformer with the same power rating. In addition, a toroid seems to draw much more current at turn-on, thus requiring a larger AC line fuse. I tend to believe that weight is an indicator of transformer quality, whether the transformer has a toroid core or an E-I core. The heavier the transformer, the better it is.

Do I have to match the transistors?

No. I never matched the transistors in any of the amplifiers I built and I never had any problems with them. If an amplifier has more than 50 mV DC offset at the output with the inputs disconnected, you may wish to check the matching of the transistors in the input diff amps.

One aspect of your design that's most curious is putting the 10 ohm - 0.1 uF series output compensation after the output inductor. In a normal network, the hf load is before the inductor. As I understand it, the whole purpose is to keep a low impedance load on the amp at high frequencies when it's driving an inductive load to help prevent spurious oscillation and other instabilities. Putting this network after the inductor reduces its ability to do this.

The first amp I built had the R/C network on the circuit board before the R/L network. The amp was subject to random overheating. I found it was oscillating. I moved the R/C network to the speaker output terminals and the problem was solved. The problem was caused by the high frequency currents in the R/C network flowing back through the circuit board ground wire, causing positive feedback into the circuit. Putting the network on the speaker outputs solves this problem, for the currents flow in the speaker ground wire instead of the circuit board ground wire. With the R/L network before the R/C network, the amp is not driving an inductive load at high frequencies because the R is in parallel with the L, not in series with it.

You do not explain the logic of using a fully complimentary input stage and voltage amplifer stage. Is it because of slew rate symmetry? Most designs, even high-end ones, use a single differential pair and a single voltage amp transistor. If the designer adds additional transistors, normally they're used for current mirrors, cascode stages or as a buffer between the voltage amp and the first emitter follower. All of these have the potential for increasing linearity while your extra parts only provide additional (and perhaps cosmetic?) symmetry.

My original intent was to design a fully complementary amplifier because the push pull action of such circuits generates less distortion. Therefore, I used complementary diff amps. Incidentally, I have seen these used in some very high performance Harris op amps. With the complementary diff amps, the positive and negtive sides of the amp turn on at the same rate, so there are no thumps in the loudspeaker, and the slew rate is perfectly symmetrical. Both diff amps are very linear. They will put out undistorted sine waves with a peak-to-peak voltage of over 4 V before clipping. This is far more voltage than what is required to drive the second stage into clipping.

The complementary or push pull voltage gain stage is superior to a single voltage amp transistor, i.e. a single ended stage, because the push pull action cancels even order distortion components. In vacuum tube days, push pull symmetry was always considered superior to single ended designs, although single ended output stages seem to be making a sentimental comeback in some tube amps.

If only a single diff amp is used, the only way to obtain a complementary second stage is to take differential outputs from the diff amp and to use a current mirror to drive the other side of the second stage. Frequency compensation of these circuits is tricky because the two signals from the outputs of the diff amp to the output of the second stage travel through paths with different amplifier configurations. Matching the gain and phase characteristics of these paths is difficult. For good stability, you usually end up with a lower open loop bandwidth and slew rate than you could obtain with complementary diff amps at the input.

In addition, I don't like to use a current mirror in a gain path. With discrete transistors, it is difficult to match the two transistors in the mirror. Even when they are well matched, the Early effect and temperature effects cause output current to be greater than the input current. Indeed, the current ratio varies with voltage across the second transistor, which varies with the signal. Series emitter resistors in the mirror are a partial fix for this problem.

You evaluated a number of output stages using SPICE and I'm rather surprised you found the triple Darlington to be the best. Best in what way? With 3 base/emitter junctions in series, triple Darlingtons tend to exhibit all sorts of non-linear behavior. Due to extremely high current gain, they're also prone to spurious oscillations (which you apparently have seen your share of). Yes they do have a high input impedance and provide high current, but you've tightly limited the current capability of the amp with the protection circuits anyway. Are you familiar with a Sziklai-Pair (Complimentary Feedback Pair)? They tend to be far more linear although they do present a somewhat lower input impedance but you also save the expense and non-linearity of a set of transistors.

You are assuming that the transistors in the driver stage operate class-AB. In the configuration that I used, only the output transistors operate class-AB. The four driver transistors operate class-A. In the Sziklai connection, both the driver transistors and the output transistors operate class-AB. This increases crossover distortion and leads to problems with parasitic oscillations.

My criterion in comparing the output stages was output resistance. A perfect voltage amplifier has zero output resistance. The T circuit version of the triple Darlington had the lowest resistance of the stages I compared, including the Sziklai connection.

I have never really experienced non-linearity problems with the triple Darlington that I could identify. The Low TIM Amp never had spurious or parasitic oscillations. However, I did experience them in the Double Barrelled Amp. Adding series base resistors on the output transistors solved the problem. When I discovered that, I added the resistors to the Low TIM Amp for good measure.

The Sziklai connection is an interesting one, and at one time I thought it would make a better output stage. It has local series-shunt feedback which I thought would give it the lowest output resistance. However, it was not as low as the triple Darlington. This is because the Sziklai connection has a very high output resistance without its local feedback. The local feedback does not lower it to what you can achieve with the Darlington. On a further note, the conventional Sziklai connection has only two transistors in it. To make it work in a high current output stage, an emitter follower must be added between the two transistors to drive the output transistor. You end up with the same number of transistors as with the triple Darlington.

Even though the Sizklai connection operates at unity gain, it has voltatge gain that is reduced to unity by local series shunt feedback. I really dislike operating the output transistors in anything but a unity gain configuration, for that is the only way you can get respectable bandwidth out of them. In the Sziklai connection, high frequency load current must come from the driver and pre-driver transistors, making these transistors subject to failure. With the triple Darlington, all load current comes from the final output transistors, regardless of frequency.

I never did a comparison of the different output stages until well after the original amp was designed. A student did a special project where he simulated the stages with SPICE for me and found the T circuit version of the triple Darlington to have the lowest output resistance over the widest frequency band.

I really never had any of the problems you mentioned with the triple Darlington. The only times I have had problems are when students design and build an amp with a different output stage leading to problems with asymmetrical clipping and parasitic oscillations. I have seen them all.

I'm curious why you chose to use four diodes for the thermal compensation instead of just mounting the Vbe multiplier (Q7) to the heatsink? It's quite easy to achieve good thermal stability with a triple Darlington this way and it makes for very simple construction if you use a small power device (like a TO-126 or TO-220) for Q7 that can be readily mounted to the heatsink.

I like the idea of having the Vbe multiplier on the circuit board to minimize stray capacitance to ground from the output of the second stage. This leads to better stability. Remember, the second stage sets the dominant pole in the open-loop transfer function. I don't want that transfer function to be a function of unpredictable wiring capacitance to a Vbe multiplier on the heat sink. Therefore, I opted for the diodes on the heat sink. The stray capacitance of the leads to the diodes can be isolated with series resistors. You can't use the series resistors if you put the Vbe multiplier on the heat sink.

The original amp used 2 diodes. I changed to 3 to after measuring the bias current as a function of heat sink temperature and found the amp to be thermally undercompensated. But with 3 diodes, the two leads from the diodes to the circuit board come out on opposite sides of the heat sink. I added a 4th to recitfy this problem and found that the amp didn't seem to be thermally overcompensated.

Did you run any open-loop measurements (or even simulations) on the amp? I'm curious what the open loop performance was like and how much feedback you're running at say 1khz and 20khz. You call it a "low feedback" design but you don't specify the parameters. While the input degeneration probably helps, I think the open-loop linearity is severely hurt by the lack of current sources for both the input and voltage gain stages and by using a triple output stage.

With complementary diff amps, it is impossible to use current source loads on the inut stage. The bias currents in the two diff amps are not stable. You must use resistive loads. For frequencies above the first pole frequency in the open loop amplifier, most of the signal output current from each diff amp flows in the compensation capacitor in the second stage, so current source loads in the diff amps would not result in an improvement above the first pole frequency. If current source loads were added, the first pole frequency would move down with an attendant increase in open loop gain. This would increase the amount of feedback at the lower frequencies. With the resistive loads, it is already high enough at the low frequencies so that I do not feel that any improvement would result. This is a moot point, however, because current source loads cannot be used on the diff amps in a fully complementary amplifier.

You don't seem to understand that each transistor in the second stage does see a current source load. With the fully complementary second stage circuit, each side of the stage sees a current source load which has both ac and dc currents in it. The dc currents are common mode. The ac currents are differential. When one is increasing, the other is decreasing, and vice versa. This push pull action cancels even order distortion components. This cancellation would not occur if only one side of the second stage is driven by the signal and the other side is a constant current load as you suggest.

My original article in Audio described the open loop measurements. It seems I had a 40 kHz open-loop bandwidth and 0.5% THD just below clipping at 1 kHz with an 8 ohm load. Later, I decreased the bandwidth while holding the gain bandwidth product constant so that the closed-loop bandwidth did not change. Since then I decreased the gain bandwidth product a little which reduced the closed-loop bandwidth. This caused a small increase in distortion because it decreased the amount of feedback. However, the stability margin is improved and I think that leads to a better sounding amp.

The first two transistors in the triple Darlingtons do not supply load current so that the signal currents in these transistors are the base input currents for the following stage. This configuration was called the T circuit by Bart Locanthi when he published it back in the 1960s. It differs from the conventional Darlington in that the resistors which set the bias currents in the first two transistors connect to the opposite transistor rather than to the amplifier output. This feature does two things. First, the driver transistors operate class A, not class AB, so that no non-linearity is caused by the switching on and off of driver transistors. Second, the output resistance is lower. I do not think that this circuit suffers from any of the problems you associate with the triple Darlington.

In comparing the amount of feedback in amplifiers, the gain-bandwidth product is a parameter that is more important than the gain. If you take an amplifier with an open loop gain of 40 dB and a bandwidth of 40 kHz, it has a gain-bandwidth product of 4 MHz. You reduce the gain to 20 dB with feedback and you have a closed-loop bandwidth of 400 kHz. With only 20 dB of feedback, it can be called a low feedback amplifier.

Now, suppose you increase the open-loop gain to 80 dB while simultaneously reducing the open-loop bandwidth to 400 Hz. With a closed-loop gain of 20 dB, you will still have a closed-loop bandwidth of 400 kHz. The amplifier is still a low feedback amplifier. In contrast, if you increase the open-loop gain to 80 dB while keeping the open-loop bandwidth at 40 kHz, you would have a closed-loop bandwidth of 4 MHz with a closed-loop gain of 20 dB. The amp is now a high feedback design and would surely oscillate like crazy.

I guess what I am saying here is that you should look at the open loop gain above the first pole frequency when you are comparing the amount of feedback. If you increase the open loop gain while simultaneously decreasing the first pole frequency, the gain above the former first pole frequency does not change.

I only have the information that's currently on your website so I'm sure I'm missing some of the background of your design. It's still intriguing, however. I learn something from nearly every design I look at (even if it's how to save money!). I appreciate your taking the time to publish yours (both in Audio and on the web).

I was pretty lucky in the original design of the amp because it worked so well. Every time I made changes, I tried to keep it as close to the original as I could. The feedforward split was added only recently. I did it simply because it made sense. I have supervised more students than I could count who built the amp, and I have experienced just about every problem that could occur. One of the worst to diagnose was when a student last fall got the lead to one of the bias diodes reversed with a lead to an output transistor. The bias in that channel would not set right. Before he found the error, he replaced the bias diodes and so many parts on the circuit board that pads started peeling up. You name it, I have seen it.

Can I use a regulated power supply with the amplifier?

Yes. I have had one student who used regulated power supplies because the bargain transformer that he bought had too high a secondary voltage rating. He designed regulators to drop the voltage down to the required values. I have no recommendations on the design of the regulators. They must be able to pass the full amplifier current and the pass transistors in them must be heat sinked. In effect, each regulator is another power amplifier that supplies voltage and current to the amplifier.

Can I strap either the Low TIM amplifier or the Double Barreled Amplifier?

I have never done this. A stereo amp can be converted into a mono strapped amplifier by driving both channels with the same signal. However, the signal to one channel must be inverted, i.e. multiplied by -1, with an op amp inverter. Strapping an amplifier doubles the available drive voltage to the loudspeaker, thus quadrupling the available power. However, each channel of the strapped amplifier sees one-half the loudspeaker impedance. I know of no reason why either amplifier can't be strapped, but I have never tried it. I would not recommend strapping any amplifier to drive low impedance loudspeakers.

This page is not a publication of the Georgia Institute of Technology and the Georgia Institute of Technology has not edited or examined the content. The author of this page is solely responsible for the content.