![\"Figure](\"https:\/\/tomorrow82.ru\/wp-content\/uploads\/2020\/03\/fig1.gif\") <\/td><\/tr> | ||||||||
| Figure 1. (a) Amplifier block diagram. (b) Feedback and error voltage waveforms.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n <\/p>\n\n\n\n Suppose that the input stage to the amplifier modeled in Figure 1(a) \nis a conventional BJT diff amp. One BJT in a diff amp cuts off when the \ndifferential input voltage exceeds about 57 mV. The differential input \nvoltage in Figure 1(a) is the error voltage ve<\/sub><\/em>, so that the diff amp would overload for ve<\/sub><\/em>\n = 0.057 V. It follows from Figure 1(b) that a 1 V step at the input \nwould cause gross overload of the diff amp. There are two ways of \nincreasing the overload voltage. One is to use resistors in series with \nthe emitters of the BJTs. The other is to replace the BJTs with FETs. \nBecause FET parameters are so unpredictable and FETs are more \nsusceptible to flicker noise, I prefer the BJTs with emitter resistors. \nThis is explained in more detail below.<\/p>\n\n\n\n The time delay before the feedback signal arrives at the input stage \nis set by the bandwidth of the open-loop amplifier. TIM cannot occur if \nthe open-loop bandwidth is greater than the signal bandwidth. Thus a \ncriterion that has been described for minimizing TIM is to design the \nopen-loop amplifier to have a bandwidth greater than the audio band. To \nkeep the amplifier stable, the product of the open-loop gain and the \nopen-loop bandwidth must not be too large. Thus if the open-loop \nbandwidth is increased, the open-loop gain must simultaneously be \ndecreased to keep the amplifier stable. With a low open-loop gain, such \ndesigns were said to be «low feedback amplifiers.» I published an \namplifier in Audio<\/em> magazine which had an open-loop bandwidth of \nabout 25 kHz and an open-loop gain of about 400. I called the article \n«Build a Low TIM Amplifier.» With no feedback, the amplifier had a THD \nat 1 kHz of 0.5% at an output power of 50 W into 8 ohms. <\/p>\n\n\n\n Many of the critics of TIM disputed the design criteria that the \nopen-loop bandwidth must be greater than the signal bandwidth to prevent\n TIM. After my article was published, I came to realize that this is \ntrue, provided the open-loop gain and bandwidth are varied in such a way\n that the product of the two remains constant. If this is done \ncorrectly, static distortions such as THD and IM can be reduced while \nnot affecting the stability of the amplifier or its susceptibility to \nTIM. The product of the open-loop gain and bandwidth is called the gain \nbandwidth product. The original amplifier had a gain bandwidth product \nof 10 MHz. I have since decreased this to about 8 MHz because I feel \nthat it improves stability and leads to a better sounding amplifier. The\n gain bandwidth product was decreased principally by increasing the \nvalue of the diff amp emitter resistors from 100 ohms to 300 ohms. This \nalone would have decreased the gain bandwidth product to below 8 MHz, \nbut I simultaneously changed the transistors in the second stage to \nlower capacitance transistors which tended to increase it. <\/p>\n\n\n\n I could not even guess how many people built the original amplifier \nor the several versions that followed. I received mail from all around \nthe world. It seemed to me that the interest in building it would never \nstop. By the early ’90s, however, about the only people I heard from who\n wanted to build it were students. Then a former student sent me email \nand asked if I had my amplifier plans on the web. After I put a very \nbrief set of plans on the web, I started getting email from people \ninterested in the amplifier. The interest seemed to peak up when Georgia\n Tech hosted the «Olympic Village» during the 1996 Olympics in Atlanta. \nEvery time I have received questions, I have tried to modify the plans \nto answer them. Thus the plans have gotten much more detailed than I had\n originally intended. <\/p>\n\n\n\n The Leach Amp, Vers. 4.5, is the latest update of the «Low TIM \nAmplifier.» The Georgia Tech students who built it never referred to it \nas the «Low TIM Amplifier.» They always called it «The Leach Amp.» Over \nthe years, I have seen countless versions of the amplifier built by \nstudents and others. All of the bugs have been worked out, and the \namplifier should perform flawlessly if it is built with patience and \ncare. This is an advanced construction project. I do not recommend that \nsomeone tackle it who has not had experience with electronic \nconstruction and assembly. If you don’t know how to solder electronic \ncomponents to a printed circuit board, this is not a project that I \nrecommend. I recently had a student solder the components to his circuit\n boards and test them in one of our labs. He told me that they tested \nok. I asked to see the boards. On one, the base lead to a transistor was\n not soldered. On the other, there was a cold solder joint through which\n the component lead could be pushed out of the hole in the board. If \nthese had been left uncorrected, neither channel of his amp would have \nworked after connecting the output transistors to the circuit boards. \nYou must know how to solder to tackle this project.<\/p>\n\n\n\n There was a mechanical engineering student at a school in New York \nstate who tried to build an amplifier. I got several emails from him \nasking questions while he was building it. When he finished, he plugged \nthe amp into the wall outlet, turned it on, and it flamed. I felt really\n sorry for the guy. His error was not double checking everything he did \nand he did not follow the test procedures that I described. One must \ncheck the circuit boards with a current-limited bench power supply \nbefore and after connecting the power transistors to them. Lacking a \nbench power supply, the amplifier power supply with a Variac transformer\n to vary the ac voltage can be used, but I don’t recommend it. In this \ncase, the circuit board fuses should be replaced with 100 ohm 1\/4 watt \nresistors. If there is an error, these will smoke and protect the \namplifier. This is all described in the Construction Details. I heard \nfrom one builder who said that his amplifier would have gone up in smoke\n if he hadn’t followed my advice about the 100 ohm resistors. I never \nheard again from the student in New York after he said that his \namplifier blew up.<\/p>\n\n\n\n When students ask me how long it takes to build the amplifier, I tell\n them to allot 3 months to order and acquire the parts and a second 3 \nmonths to build it. A former student thought he could build the \namplifier in two weeks. Not only did he pay local vendors about twice \nwhat he could have ordered the parts for, but his amplifier was plagued \nwith construction problems. In trying to solve the problems, he \nunsoldered and soldered parts to the circuit boards so many times that \nthe traces started to peel up. He ended up having to get a new set of \ncircuit boards and start over. After he graduated, he has ordered eight \nmore boards to build more amplifiers. <\/p>\n\n\n\n With power supply voltages of about 58 V dc (+ and -), the amplifier \nwill put out an average sine wave power of 120 watts per channel with an\n 8 ohm load. SMPTE IM distortion is typically 0.15% at clipping and less\n than 0.01% at lower levels. The circuit has a voltage gain bandwidth \nproduct of about 8.5 MHz and a large signal slew rate of 60 V per \nmicrosecond. Frequency compensation is provided by 300 ohm emitter \ndegeneration resistors in the input differential amplifier stage, 10 pF \nlag compensation capacitors in the high gain stage, and feedforward \ncompensation above 150 kHz around the driver and output stages. Voltage \nand current sensing protection circuits prevent damage to the amplifier \nin the event that the output is short circuited. <\/p>\n\n\n\n Some readers have questioned the 8.5 MHz gain bandwith product. This is not to be confused with the fT<\/sub><\/em> or gain bandwidth product of the output transistors. The fT<\/sub><\/em>\n of a transistor is the frequency at which its current gain is reduced \nto unity. In contrast, the gain bandwidth product of an amplifier is the\n product of the voltage gain and the frequency at which the gain is down\n by 3 dB or a factor of 0.707. Regardless of the amplifier circuit, this\n product cannot be made too large if the amplifier is to be stable and \nnot oscillate. Setting the gain bandwidth product is part of what is \ncalled the «frequency compensation» of an amplifier.<\/p>\n\n\n\n The amplifier will drive a 4 ohm load to full power without current \nlimiting. Depending on the power supply regulation, the output power \nwith a 4 ohm load is as great as twice the power with an 8 ohm load. \nWith loads lower than 2 ohms, the protection circuits limit the maximum \noutput current, and thus the output power, to protect the output \ntransistors. The amplifier is stable with capacitive loads and will \ndrive electrostatic loudspeakers with no problems. <\/p>\n\n\n\n A good indicator of an amplifier sound quality is its ability to \ndrive a capacitive load. The classic example of a capacitive loudspeaker\n transducer is the electrostatic transducer. Another is the \npiezoelectric transducer used in some horn tweeters. I have heard of \namplifiers overheating when driving arrays of these tweeters. No doubt \nthe circuits were oscillating. A third source of load capacitance that \nis often overlooked is the loudspeaker cable. Some of the so-called \n«high definition» cables are designed to minimize the series inductance.\n Because the series inductance per unit length multiplied by the shunt \ncapacitance per unit length is a constant that is equal to the \nreciprocal of the velocity of light squared, minimizing the inductance \nmaximizes the capacitance. Therefore, these cables can exhibit a high \nshunt capacitance. For this reason, I generally do not recommend them. I\n have heard of a certain «high-end» amplifier that is no longer made \nsmoking when these cables are connected to its output, even with no \nsignal input. A former student, who was working for Panasonic, performed\n a listening test in their «high-end» listening room to see if there was\n any audible difference between a «high definition» cable and ordinary \nzip cord. I do not know what power amplifier they used, but he said that\n the listeners were in agreement that the high-frequency response was \nbetter with the zip cord, no doubt because the high capacitance of the \n«high definition» cable.<\/p>\n\n\n\n The Low TIM amplifier is stable with capacitive loads. I have tested \nit with a 2 uF capacitor. This is a test that I learned from reading \nBascom King’s reports on amplifiers for Audio<\/em> (rip) magazine \nback in the ’70s, and it is probably the worst load that I know of. \nBecause some amplifiers become unstable with a capacitor for a load, I \nhave seen the test made with a resistor, e.g. a 2 ohm resistor, in \nseries with the capacitor. This is cheating. It does not indicate the \nstability of the amplifier with a capacitive load. The first time I \ntested my prototype Low TIM amplifier with a 2 uF capacitor, it blew the\n 0.33 ohm 5 W emitter degeneration resistors in series with the output \ntransistors. I suspect the circuit was oscillating. Either the resistors\n couldn’t handle the current or the circuit was oscillating. I suspect \nthe latter. These problems were solved before I finished the original \nprototype amplifeir.<\/p>\n\n\n\n Another good indicator of an amplifier sound quality is its high \nfrequency clipping behavior. When an amplifier is driven into clipping, \nsome of the transistors in the circuit are driven into saturation. When a\n transistor saturates, a relatively large charge is stored in the \nsemiconductor junctions. For the amplifier to come out of clipping, this\n charge must be neutralized. To minimize the time for this to occur, the\n larger output and driver transistors should never be allowed to \nsaturate. Although the power rating of an amplifier can usually be \nincreased if the output transistors are allowed to saturate, it causes a\n high frequency clipping problem that has been described as «sticking.» \nWhen an amplifier exhibits this problem, its output waveform appears to \nbecome «stuck» at the clipping level when it is driven into clipping at \nhigh frequencies. In the Low TIM amplifier, the transistors which \nsaturate are in the second stage of the amplifier. As a result, sticking\n problems are minimized when the amplifier is driven into clipping.<\/p>\n\n\n\n The data displayed in Figures 2 through 5 were measured on student \nAllen Robinson’s amplifier with an Audio Precision System II analyzer. \nFigure 2 shows the plot of the measured gain versus frequency over the \nband from 20 Hz to 20 kHz. Below 10 kHz, the gain is 26.6 dB, decreasing\n to 26.5 dB at 20 kHz. The gain was measured with an input voltage of 1 V\n rms with an 8 ohm load. At this level, the output power is 57 W.<\/p>\n\n\n\n Figure 3 shows the plot of the measured phase versus frequency over \nthe band from 20 Hz to 20 kHz. As with the gain measurement, the phase \nwas measured with a 1 V rms input signal and an 8 ohm load. At 20 Hz, \nthe phase is +1.5 degrees, decreasing to -8.8 degrees at 20 kHz. If a \nlinear scale were used for the frequency axis, the phase plot would be a\n line with a constant slope. The slope corresponds to the time delay \nthrough the amplifier which is approximately 1.5 microseconds.<\/p>\n\n\n\n Figure 4 shows the measured total harmonic distortion versus \nfrequency at an output power of 120 watts into 8 ohms. At 20 Hz, the \ndistortion is 0.006%. It increases to 0.29 % at 20 kHz. The rise in \ndistortion at higher frequencies is a characteristic of a low feedback \namplifier. It is caused by the decrease in loop gain above the first \npole frequency in the amplifier transfer function. The loop gain \ndecreases at 20 dB per decade above the first pole, causing the \ndistortion curve to increase at high frequencies. The feedforward \ncompensation around the output stage at high frequencies also \ncontributes a little to the rise. At lower power levels, the distortion \nis less, especially at the higher frequencies.<\/p>\n\n\n\n Figure 5 shows the measured intermodulation distortion versus input \nlevel with an 8 ohm load. The distortion was measured using the Society \nof Motion Picture and Television Engineers (SMPTE) standard. The input \nsignal is the sum of a 60 Hz sine wave and a 4 kHz sine wave having an \namplitude ratio of 4 to 1. The analyzer measures the percent modulation \ndistortion on the smaller amplitude 4 kHz wave caused by the larger \namplitude 60 Hz wave. The input voltage was swept from 0.1 V peak to 2 V\n peak for the measurements. The amplifier is at the threshold of \nclipping when the input voltage is 2 V peak. At this level, the IMD is \n0.018%. At lower levels, it decreases to 0.0059% and then increases as \nthe level is decreased further. The rise in IMD at the lower levels is \ncaused by residual noise and hum in the measurement system. Although the\n hum and noise is constant, it increases when expressed as a percent of \nthe output voltage when the input voltage is decreased.<\/p>\n\n\n\n Since making these measurements, I have discovered that the Audio \nPrecision analyzer puts out a constant low-level noise that gets \namplified by the amplifier and causes the distortion to measure higher \nthan it should at low power levels. No doubt, this is the reason for the\n rise in IM distortion at low levels in Figure 5. A student discovered \nthis in testing his amplifier. The solution he found was to increase the\n output level of the analyzer and use an attenuator between it and the \namplifier. I need to retest Allen’s amplifier with the attenuator. The \ndistortion at low levels should be lower that what is shown in the \ngraph.<\/p>\n\n\n\n A hot topic of amplifier design in the 1970s was «transient intermodulation distortion» (TIM). Other names which were used for this phenomenon were «slewing induced distortion» (SID), and «dynamic intermodulation distortion» (DIM). TIM occurs when a transient input signal overloads the input stage of an amplifier, causing it to either cut off or to become… <\/p>\n |
![\"Figure](\"https:\/\/tomorrow82.ru\/wp-content\/uploads\/2020\/03\/mag.gif\")
<\/td><\/tr>![\"Figure](\"https:\/\/tomorrow82.ru\/wp-content\/uploads\/2020\/03\/phase.gif\")
<\/td><\/tr>![\"Figure](\"https:\/\/tomorrow82.ru\/wp-content\/uploads\/2020\/03\/thd.gif\")
<\/td><\/tr>![\"Figure](\"https:\/\/tomorrow82.ru\/wp-content\/uploads\/2020\/03\/smpteim.gif\")
<\/td><\/tr>