THE STORY BEHIND the A75 power amplifier continues. In Part I, we discussed its origins and direct predecessor, the Pass A40, as well as some general guidelines to follow before initiating such a project. We covered such topics as MOSFETs and differential pairs. In this concluding part, we will analyze… More...

THE STORY BEHIND the A75 power amplifier continues. In Part I, we discussed its origins and direct predecessor, the Pass A40, as well as some general guidelines to follow before initiating such a project. We covered such topics as MOSFETs and differential pairs. In this concluding part, we will analyze the amplifier's component parts in depth, beginning with the source.

The Power Supply

Figure 11 shows the amplifier's power supply circuitry. AC line power enters through the power cord and passes through fuse F1. To reduce the effects of inrush current, thermistor TH1 is employed. At room temperature, this will have a resistance of several ohms, which will limit the initial power supply capacitor charging current. A short while after start-up, the thermistor receives heat from the current passing through it and has a low impedance.

Following the thermistor, we see spike absorber TZ1 which conducts at high spike voltages and which, in this application, protects the triac. In parallel with it is C1, which aids in spike suppression and RF filtering. Triac TR1 is the main power switch. Rated at 40A and 600V, it is capable of withstanding high inrush surges. You can easily trigger it with small switch S1, which is in series with limiting resistor R2. Across the triac is an RC network formed by R1 and C2, which damps out transients across the triac. C3 follows the triac and supplies more suppression and filtration.

The transformer is a 500W toroid with dual 115V primary and 30V AC secondary coils. Figures 11 and fig. 16 show the hookup of the primary coils for both 115V and 230V operation.

The secondary system is comprised of a standard unregulated plus and minus 37V DC supply and a higher voltage regulated supply. The high power unregulated supply drives the output devices of both channels and consists of BR1, BR2, and four large capacitors, C10-13. Two bridges are employed to reduce the work load, and to afford some isolation between the two power supply channels. These supplies will communicate with each other only during the charge pulse which occurs at the peaks of the AC line waveform.

The regulated high voltage supply will drive the front ends of the amplifier channels, and delivers a regulated +/50V at about 150mA. It consists of C4-7 and D1-4, and is formed by a voltage doubler circuit. Without a load, this circuit will double the supply voltage of +/- 35V of the unregulated supply, but without much current capacity. Resistors R3 and R4 slow the charge cycle and reduce the output voltage, which lowers the heat dissipation in the regulator circuits.

Voltage Doubler

Voltage doubler operation is not easily explained, and sometimes you just have to look at it for a while. Since the two halves are independent,

let's look at only the positive half. When transformer tap A swings negative, it will cause C4 to pick up a charge of equal voltage as conducted through D1 from ground. Since tap A will hit about - 38V relative to ground, the capacitor will be charged to approximately 36V.

When tap A swings positive, the voltage on the capacitor will swing with it, plus the 36V at which it has been charged. When tap A reaches +37V, the voltage on the capacitor's positive side will now be at 37V + 36V = 73V. This 73V will conduct through diode D2 and charge capacitor C6 up to about 72V. The result of both the positive and negative voltage doubler circuits is + 72V and - 72V, but is poorly regulated with large ripple voltage.

The remaining circuitry consists mostly of the plus and minus regulators for the high voltage front end supply. On the positive side, R14 sends current through Z2 to create a 9.1V source. This biases up the differential pair formed by Q7 and Q6 through R11. With an input voltage at 9.1, the voltage across R11 will be about 8.4V or 4mA.

Q7's collector output drives the base of Q3, forming another example of our friend, the op amp. Q3's output is fed back to the base of Q6 (the op amp minus input) through resistor network R10 and R13. These form a voltage gain of 5.55, which multiplied by the input voltage of 9.1 gives an output voltage of 50.5V. Capacitor C8 provides some filtering and assures the loop stability of our op amp when used as a power supply regulator. Note that R11 is held at a constant voltage by the constant voltage input, eliminating the need for a constant current source to improve the performance.

Cooling for the output stage heatsinks is provided by a DC fan. You can control its speed very easily by varying the DC voltage. Generally, running these fans at full voltage will create much more noise than the audiophile will wish to hear. Our fan will be run slightly under 15V (although it is rated at 24V), which will be enough to reliably push air through the heatsinks but will not be too noisy.

The fan voltage is derived by running the positive regulated 50V supply through resistive dividing network R19 and R18. This reduced voltage is

their fairly low capacitance and greater linearity at lower currents compared with higher current types.

Q3, Q9 and Q7 (and their complements Q6, Q10 and Q8) operate at higher currents. Q9 and Q10 also operate at significantly higher wattage, and a larger transistor was appropriate. We chose IRF510 and its complement IRF9510. The output devices were chosen with similar criteria. The high-performance IRF230 costs as little as $2 each in quantities.

For all these devices, there is some latitude in alternate parts. When you call the distributor, it is unlikely he will have all of them in stock (although one of the reasons you see them here is that they were available when the amp was designed). Here is a list of alternatives from International Rectifier.

IRFD110: N-channel, 1W, O.5A, RDS < 1ohm
Examples: IRFD113, IRFD123, IRFD120, IRFD223, IRFD210, IRFD220

IRFD9110: P-channel, 1W, 0.5A, RDS < 1ohm
Examples: IRFD9113, IRFD9123, IRFD9120, IRFD9220

IRF510: N-channel, 20W, 4A, RDS < 1ohm
Examples: IRF512, IRF612, IRF610, IRF710, IRF712

IRF9510: P-channel, 20W, 4A, RDS < 1ohm
Examples:IRF9512, IRF9612, IRF9610

IRF230: N-channel, 75W, 9A, RDS < 1ohm
Examples: IRF130, 2N6756, IRF231, IRF232, IRF233, IRF230, N6758, IRF330

IRF9231: P-channel, 75W, 9A, RDS < 1ohm
Examples: IRF9230, IRF9232, IRF9233, IRF9130, IRF9132

Any of these devices would serve as equivalent replacements. They

Consistency is the most important thing here. The given voltage is 15 and, adjusting for about a 4V VGS, we will see about 11V across the resistor.

We are looking for as much matching of the input MOSFETs as possible at a current of 5mA. For this test, we use an R1 value of 2.2kohm. Measure the voltage between the gate and the source. Write it down on a piece of masking tape or a sticky label and place it on the part. Keep in mind the caveats about electrostatic discharge: touch ground before you touch the parts.

Matching input MOSFETs is critical, because they must share equally the 10mA of bias current from the current source, and they will not do that unless their VGS is matched. At 5mA current, they have an equivalent source resistance of about 15ohm. Assuming we want them to share the current to within 2mA, we calculate the required VGS match as follows. Using the formula V = IR, we see V = 0.002 x 15, which gives us 30mV. The VGS of the input devices should be matched to within 30mV at 5mA current. The matching is only essential within a given pair; you do not have to match the Ps to the Ns, or match to devices in another channel.

If you are unable to find input devices matched to within 30mV, you must insert resistance in the source to make up the difference. The resistance is calculated by the difference of the two values of VGS divided by 5mA. For example, if the difference in VP1GS is 100mV, then 0.1/0.005 = 20ohm. You would then place 20ohm in series with the MOSFET source having the lower VGS.

We use the same test setup for the MOSFETs in the TO-220 packages but at a higher current (20mA), so we use a 560ohm resistor. No matching is required for these devices; we are just checking to see that the VGS is between 4-4.6V and that they work.

We will measure the output device VGS at about 170mA. You can achieve this with either a 56ohm at 2W resistor, or two 100ohm at 1W resistors in parallel. We are looking to obtain a reasonable match within a parallel output bank of each polarity of each channel, so we want two groups of 12 with matched N- channel devices, and two groups of matched P-channel devices.

The VGS voltages of our test samples gave the following spread:

------------N-channel ---- P-channel
Min. VGS ---- 4.00V ------ 3.79V
Max. VGS --- 4.57V ------ 4.15V
Avg. VGS ----4.42V ------ 4.01V

We also measured the transconductance by taking another reading for each device at a higher current (0.5A), just to see what kind of variation we got. The transconductances measured from a low of 1.19 to a high of 1.56, with the average at about 1.35. Within this amplifier's general operating curve, each output will vary its current by about 1.3A for every volt of its VGS change. For 12 devices in parallel, we expect about 15A for each such volt.

By placing 1ohm source resistors on each transistor, we can assure

hundred of you will write to ask how to bias for Class A into 1ohm.

Let's review a couple of things. The peak voltage on a clean output sine wave is 1.414 times the average voltage, and the peak wattage is twice the average. An amplifier that delivers 75W into 8ohm average will need to deliver a peak of 150W, which requires a peak voltage of 34.6: V = SQRT(150X8).

This 34.6V peak will mean that, allowing for 2V loss, you need a supply voltage on the output stage of at least 37V. A transformer having split 28V secondaries will generate this. These secondaries will peak at 28 x 1.414 = 40V, with at least 1V lost to diodes and ripple on the capacitors. The transformer secondary voltage will vary with load; the specified transformer will load down to about 37V DC when bias is applied.

The peak current from that 34.6V means peak amperage into 8ohm of 33.5/8 = 4.3A. On a push/pull Class A amplifier, the peak output current in Class A mode is equal to twice the idle current, which will be 2.2A. Into lower impedances, it will be higher.

Now 2.2A through a bank of transistors having 37V across it comes out to V x A = 2.2 x 37 = 81W on the positive bank. You will have another 81W across the negative bank, for a total idling dissipation of 162W. Again, this will be slightly more than twice the rated 8ohm power. If you want to run the amplifier at 150W Class A into 4ohm, you can do it, but you will be idling at 4.4A or 324W. You would need a larger transformer.

By the way, the transformer is selected very simply: it should have the appropriate secondary voltage at twice the VA rating of what you will use. Maybe a little less if there is a fan playing air across it. For this amplifier, we will be running about 330W and using a 550W transformer and a fan.

Anyway, here we are with 330W to dissipate in our heatsinks, and we have made it a rule of thumb not to exceed 55°C on a heatsink. Human skin has the remarkable characteristic that we think 40° is comfortable, 45° is hot, 50° is very hot, and 55° is untouchable. This expanded

Pass recommends the book Zen and the Art of Motorcycle Maintenance, by Robert Pirsig.

PC Boards

We will not go into PC board construction here, but are presenting artwork and making disk files of Gerber artwork available for those of you who want to use a PC fabrication house. These shops once accepted paper or taped layout artwork and would create the appropriate film pattern. Now it seems that they all require Gerber files exclusively.

The artwork presented here is for the front end and power supply boards. The output stage board is not shown on the presumption that you will probably end up with a slightly different heatsink layout. However, the pattern is included with the Gerber files, if you want it. An

fed to the base of emitter follower transistor Q1 whose collector is attached to the positive unregulated supply (so as not to load the regulated supply, and for lower dissipation), and whose emitter drives the fan's positive lead. The negative lead goes to ground. Be prepared to adjust the values of R19 and R18 to get the fan speed you want. You may be tempted to run it very slowly for low noise, but be certain that it always starts rotating on turn-on. Of course, you may decide to simply use very large heatsinks and dispense with the fan.

TABLE 1

Power Supply Parts List

Component Selection

You will find resistor and capacitor specifications in Table 1, along with some Digi-Key part numbers. Use as high quality parts as you like, but when considering substitutes be aware of how well they will fit on the PC board.

Gain device selection presents a narrower range of alternatives. We have chosen to use International Rectifier MOSFETs exclusively, because of their consistently high quality, moderate cost and lead time, and US manufacture.

For the input devices, we chose the IRFD110 and IRFD9110. They worked better than anything else from the catalog, mostly because of

have the same wattage ratings, more than adequate voltage ratings, and similar trans-conductance figures, but they will have different costs and availability. Other devices not mentioned here will also work well, as long as they have similar characteristics. Some of them are secondsourced by Motorola, and you can also consider devices from Toshiba and Hitachi. It will be helpful, but not essential, if they offer the same cases and pinouts. We have deliberately chosen the output transistors in TO-3 packages, but there is a large equivalent selection in plastic packages. You should not be afraid to use them. Because of the audio circuit's simplicity, its design is very forgiving of substitutions. The primary criteria are adequate voltage, current, and dissipation ratings.

When making substitutions, you should get one type or another. Mixing different devices together will not make matching easier. For that matter, it is generally helpful if they have the same lot codes, which means that they were made together and will tend to have similar characteristics.

We recommend buying more devices than you think you'll need. Even if you don't experience failures, you may find during testing that some are sufficiently different so that you don't want to use them. Having a sizable population is a real help if you are planning on selecting and matching.

Component Testing

After you acquire the devices, you will need to test them. You might consider running lots of tests on these transistors, but only one is essential: measuring gate-source voltage versus current. The greatest variations occur here, and it is necessary to do some matching to get proper performance. This test will also tell you whether or not the device is broken.

The test is simple and requires a power supply, a resistor, and a DC voltmeter. Figure 12 shows the test hookup for N- and P-channel types. The supply source resistance (R1) is nominal, and is found from

adequate current sharing for a fairly wide range of VGS. In Class A bias, we will be operating at about 200mA/device, which will place 0.2V across each source resistor. A variation in VGS will cause the bias to be unequally distributed between the devices. For example, for a 4.6V device in parallel with a 4.5V device, the first will run at about 160mA at 6W and the second at about 240mA at 9W.

Remember that each of these devices is rated at 75W on a cold heatsink, and maybe 50W on a hot sink. We are only going to bias them to about 8W each, so they're not going to break from a little unequal distribution. Nevertheless, we like to see the load shared, and recommend that you group the outputs by VGS as closely as possible. Matching within 0.2V will work, and O.1V is even better. Within a population of 150 transistors, you can easily get 12 sets matched to O.1V VGS at 200mA.

Heatsinking

Heatsinks are "first-and-foremost" things. We will start by figuring out how much we really need. A 75W pure Class A push/pull amplifier will theoretically operate at 150W of idling dissipation, 170W when all is said and done. You can put more wattage into 8ohm by cranking up the supply if you pay close attention to your heatsink requirements. You will also need to consider this if you want to maintain Class A bias down to a lower impedance level. The amplifier will operate very cleanly into low impedances anyway, so this is not essential. But we know that about a

temperature sensitivity has a lot to do with injury prevention, and is also very convenient for judging whether or not heatsinking is adequate. If you can't touch it, it's too hot.

Given a 50°C heatsink in a 25° ambient temperature, the difference is 25°. We want to build a heatsink system which will only rise 25° while dissipating 330W. This gives us a thermal resistance figure of 0.08 °C/W. Look in the heatsink catalog and find one with 0.08°/W rating. (Now you know why there are so few pure Class A power amplifiers on the market).

Maybe we can use several smaller sinks, say 12 with 1 deg/W ratings. In the Thermalloy catalog you can check out #15217, which Pass designed to use in Threshold amplifiers. In 6" lengths it is about 0.6°/W, and you would only need eight of them. When you think about the number of amplifiers on the market claiming pure Class A at hundreds of watts, and compare their heatsinking to this, you might conclude that something funny is going on (and you would be right).

We chose to use four pairs of forced air sinks at 5" lengths whose cross sectional dimensions are given in Fig. 13. The surface area is about 54 in²/ inch, and it is very similar to the Aham Tor #3750 or the discontinued #6180.

This sink's rating is 1.5°/section/3", so for 5" it will be about 0.9°/W. For eight sections {four pairs), we are down to 0.1125°/W. We will run enough air through them to double their efficiency, for about 0.06°C/W of thermal resistance, which should be enough. If it isn't, we'll turn up the fan.

Under these circumstances, the output devices will be sitting on sinks at about 50° and, with dissipating 8W apiece, the case temperature will be up another 10° or so. With a case temperature of 60°, the data book on these 75W devices tells us they are rated at 55W. We will be using them at about 15% of rated capacity. This means two things: they will last a while, and you will be able to kick 70A into that 0.3ohm electrostatic load, at least on peaks.

The Chassis

Many people think building a chassis is the hardest part. Pass solved the problem by first building a machine shop, with five computer-controlled milling machines (Photo 1). That's how we got this chassis, which was carved out of aluminum slabs (Photos 2, 3, 4).

Most of you will show a lot of creativity in this area. Buying a prefab chassis is a good idea, but unless you have some experience drilling sheet metal, it will still not be easy. We suggest you check out some available hole punches. You put them in a small hole you have drilled and then crank on the bolt. They work adequately and are well worth the expense.

Maybe you should consider making friends with metal shop workers (they like big amplifiers, too), or enroll in night school machine shop. Whatever you do, keep in mind that the chassis will be the temple in which your project is housed, and do a good job. Plan it carefully, take your time, and try not to be too cheap about it. For proper perspective,

output stage board is a good idea for this project, because the MOSFET gate and source resistors are best located at the output devices, and a PC board is most convenient for holding them.

Power Supply Assembly

You've got a chassis and PC boards, and you've tested and matched the MOSFETs. Now you will need a power supply to test the front end, so let's build it first. Table 1 is the power supply parts list. Figures 14a and 14b are the PC board power supply and the component placement. Figure 16 shows it wired up in the amplifier. Not much to note here except that the artwork shows component side, not copper. Diodes D5 and D6 are tacked on the back across the terminals of C4 and C5, because they were added after a hundred boards were made. Note that transistors Q2 and Q3 will be operating at about 1.3W each, so give them adequate heatsinking.

You can mount them to the chassis beneath the PC board if you wish (taking care to insulate the tab from the chassis), or you can mount them

Less...