Part Selection Guide

Active Parts

For more details about op-amps, see the companion article, Notes on Audio Op-Amps.

These are the cascode JFETs, used to bias the op-amps into class A.

The standard parts are 2N5486 (IDSS 8-20mA) and 2N5484 (1-5mA). There are many other parts that will also work. The 2N5457 is a good substitute for the 2N5484, for instance. In Europe, the BF245 series is more readily available, but beware that you have to turn them 180 degrees from the orientation you'd use for the standard JFETs.

When picking your own JFETs, Q1's minimum IDSS should probably be around 1mA, with higher being better. Q2's minimum IDSS must be higher than Q1's maximum IDSS. In other words, it must be impossible for the IDSS of any random Q2 to be lower than the IDSS of any Q1. Also, Q2's input capacitance (CISS) should be low. (Under 10pF.)

You want each JFET's IDSS to be at least as high as the op-amp's current draw, plus enough extra for the cascode JFETs and other small current drains surrounding the op-amps. For instance, let's say your op-amp has a quiescent current drain of 5mA, and you've biased it into class A with 1mA through the cascode JFETs. Therefore, you'd need a bit more than 6mA through each Q3 at absolute minimum. The 2N5486 with a minimum IDSS of 8mA would do just fine. Or, you could measure the IDSS of some 2N5459s (range: 4-16mA) to find some that are high enough.

If you want to be able to swap different op-amps in without changing the Q3s, be sure to pick your Q3s such that they will be able to power the hungriest of your op-amps. The highest-drain op-amp you're likely to use is about 15mA, so picking Q3s with IDSS higher than this will cover almost any conceivable case.

That's the simple way to approach the power rail issues. There's a lot more to this, if you're interested.

This JFET is configured as an adjustable constant current source, along with R12 and R13. Adjusting this current source sets the bias point of the buffer.

If you can't get the recommended PN4392 from your parts distributor, look for the MPF4392. It's the same thing. Failing that, most anything with a minimum IDSS of 30mA or higher (and with the same pinout, obviously) should work fine.

These transistors are configured as current mirrors, which reflects the current of the buffer biasing CCS to drive the input transistors.

The recommended 2N5087/5088 transistors aren't easily available everywhere, but they are highly recommended unless you're up for some hard-core tweaking. A poor selection here can have drastic consequences to the amp's sound quality. With an op-amp, swapping out a poor performer is easy. Desoldering 18 TO-92s and putting a different set in is not so easy. (Speaking from experience here!)

With that warning in mind, there are some alternative complementary pairs you can try. We haven't tried these on a PPAv2 circuit board, and we don't know of anyone else who has. You're on your own. In rough order of how suitable we expect them to be, the alternatives are:

• MPS8099, MPS8599
• BC550, BC560 (reversed pinout)
• BC327, BC337 (reversed pinout)
• 2N3904, 2N3906
• PN2222, PN2907

The ones marked as having a reversed pinout can simply be installed "backwards" on the board.

These are the buffers' input transistors. These accept the input signal and drive the signal to the output transistors.

See the table above for alternate part hints.

The TLEs perform two functions: they set the voltage on the ground plane above the input traces, and they provide low impedance paths for current to flow around the input section in the face of the rail isolation JFETs. If the op-amps will share a single set of rails in your amp, you only need one TLE. If each op-amp has its own rails, you need three of them.

Passive Parts

Next, you need to know your power supply voltage. Because the C1s in the PPA go from rail to rail, their voltage rating must be higher than your power supply's output voltage. For instance, if you have a 30V supply, 25V caps would be damaged by the power supply, 35V caps are good, and 50V caps are wasteful. (For more on this topic, read my article Op-Amp Working Voltage Considerations.)

It turns out that the distance from the V- pad of one of the PPA's C1s to its neighbor's V+ pad is equal to the lead spacing of 16mm and 18mm diameter caps. The board will accommodate 4 of these larger caps if your C2s are small enough and you're not using a case where the board has to slide into rails.

Optional? Only 1 cap required. Up to 8 more can be added if you use 10mm or 12.5mm caps. Do not jumper.

Largest Part Size: Designed for 12.5mm diameter caps, but 10mm, 16mm and 18mm caps can be made to work without lead bending.

These are local bulk capacitance for the op-amps. Much of the C1 discussion applies to C4 as well.

The C4s go from ground to rail, so as long as the amp is functioning properly you can get away with caps with half the voltage tolerance as the power supply's total voltage. However, it's safer to use caps with a voltage tolerance at least as high as the voltage the op-amp sees. Since you don't need much capacitance here (100 µF or so per rail), there's not much point in skimping on voltage tolerances.

The C4s probably shouldn't be much larger than 220 µF since they must be charged through the current-limiting components Q3s and R8 while the amp is powering up. You want these caps to charge up quickly when the amp turns on. Since the op-amps only "sip" a tiny amount of dynamic current, there's no benefit to making C4 large.

Optional? No.

Largest Part Size: 10mm diameter

This is for tuning the bass boost. Click that link to learn how changing its value changes the behavior of the bass boost.

In PPA version 2, we made this position much larger, so there's no longer any excuse for using low-quality capacitors here. It's directly in the signal path, so you should definitely be using polypropylenes or other high-quality linear types here.

If you're not very concerned about bass boost quality, you can just use any random metallized polyester box cap here. There's a reasonable argument for this: after all, bass boost is already a nonlinear adjustment to the sound, so how bad could it be to use a nonlinear capacitor to effect the bass boost? Personally, I think there's some merit to this argument, but since good polypropylenes really aren't all that expensive, I don't see a good reason to give into such hair-splitting.

Optional? Yes. Do not jumper.

Largest Part Size: 16mm × 8mm

A diode has a small drop across it. If your amp is wall-powered, this voltage drop isn't a problem because you can pick the wall supply's voltage to be high enough that this drop is insignificant. If your amp is battery-powered, it's not so easy to say "use more batteries", but in fact you have little choice in the matter: D1 on the amp board and D2 on the battery board form a diode OR bridge, which ensures that only the supply with the higher voltage powers the amp. Since the wall supply should always be higher in voltage than the battery pack's voltage, the amp runs from the wall supply when it's available and from the batteries otherwise.

Optional? Technically, yes, but you'd better have a very good reason to jumper across it.

R1 interacts with R2 to form a voltage divider. If R1 is much smaller than R2, this effect is negligible, which is the way you'll almost certainly want it. I imagine someone might choose to configure this to divide the voltage down by a significant amount on purpose, but that's not the intent of this layout.

Optional? Technically yes, you can jumper it, but you should put a resistor here.

Together, these rules mean that a 100 KΩ volume control is the highest value you should tolerate. See below for more information on choosing a volume control.

Optional? No.

These are the feedback resistors, which set the amplifier topology and the gain. Because the PPA is a high-end amp, the only topology we support is the Jung multiloop topology, which uses all four resistors.

The values given on the schematic are good for most purposes.

The only value you're likely to need to change is R4, to adjust the gain. You could instead adjust R3, but this would upset the impedance balance at the inputs of the op-amp, increasing distortion.

If you were to use a slow or cheap op-amp, you might want to change R5 and R6, but that goes against the nature of the PPA; you should build a PIMETA if you want to use such op-amps.

Optional? If you want a "true" PPA, no.

This resistor is for tuning the bass boost. It is an alternative to using a bass boost adjustment potentiometer. Click that link to learn how changing its value changes the behavior of the bass boost.

This resistor forms an RC-low pass filter for the op-amp power rails in conjunction with C4||C5. This attenuates any high-frequency noise on the rails, which increases stability of the op-amps. This filter opposes the drop in the op-amp's PSRR as frequency goes up, giving an overall better PSRR to the amplifier.

The value of this resistor is not critical, but 10 Ω is a good value for most purposes. This gives a corner frequency of 16 kHz with a 1 µF C5. This counteracts the op-amp's falling PSRR vs. frequency. If you want a lower corner frequency, you must raise the value of C5. Theoretically you could instead raise the value of R8, but as its value rises, so also does the current-modulated rail ripple due to the op-amp's varying current draw. We decided during development that 10 Ω was the best trade-off here.

Optional? Yes. You can jumper it.

This is the "source resistor" for the op-amp's class A biasing cascode. See this section for details of how this works.

R9 is a multiturn trim pot configured as a variable resistor. A trimmer in the 1 KΩ to 10 KΩ range will be about right. The right value depends on how much adjustment range you need, and how fine your control in setting particular values needs to be. If you don't want to experiment with this, 2K is a reasonable value.

Optional? Yes. Leave it out if you're not biasing the op-amps.

If you bias the op-amp into class A with JFET cascodes, you should also add R10. JFETs have an input capacitance, and you don't want a capacitor across the output of the op-amp. R10 is a kind of "insulation" between the two, so they don't affect each other as much.

The exact value of this resistor is not critical, but it does have to fall within a certain range. If it's too low, it doesn't do its job very well; 100 Ω is probably the smallest value that will provide sufficient benefit. If it's too high, it will cause the cascode to fall out of regulation during parts of the swing of the op-amp's output, negating its advantages; 1 KΩ is about the largest you can get away with. Since you probably are using 1K's elsewhere in the circuit, it's simplest to just double up on some of those when ordering parts.

Optional? Populate it if you bias the op-amp into class A. Leave it empty otherwise.

This trimmer is part of a constant current source — along with Q4 and R13 — which you adjust to set the output stage's bias point.

For most purposes, 2 KΩ will be fine. You might choose to go with a 5 KΩ unit instead if you need extreme adjustments, such as when you are experimenting with the buffer.

Optional? No.

This resistor simply sets a minimum on the buffer bias adjustment. You want some minimum amount of resistance here for best current regulation through Q4.

Optional? No.

This is the power indicator LED's current limiting resistor. You use it instead of the LED cut-off circuit if your amp doesn't use batteries, or you just want a simpler LED configuration. It can be a 5% carbon type; the exact value isn't at all critical.

where:

V_s is the power supply voltage, rail to rail
V_f is the LED's forward voltage drop
I_f is the desired current through the LED

1 mA gives enough brightness for a power indicator with most LEDs, but some may require a bit more. The limit for most LEDs is 20 to 40 mA, but even approaching that much current makes the LED too bright for my taste.

Typical values for RLED are 1 KΩ to 10 KΩ, depending on the power supply voltage and the LED being used.

Optional? Add RLED if you don't use the LED cut-off circuit. Leave it out otherwise.

This subcircuit is an alternative to RLED. It works best with battery powered amps that have enough supply voltage that they are able to completely drain the battery pack before they start clipping with your headphones. If this is not your situation, using RLED probably makes more sense. (If you're not sure which situation pertains, read the article Op-Amp Working Voltage Considerations.)

Rechargeable batteries drop quickly in voltage from their fully-charged state, then stay near 1.2V per cell for most of their useful life, then drop off in voltage quickly as they deplete the last of their charge. If you use RLED in this situation, the LED will vary in brightness as the batteries drop in voltage. You might be able to look at the LED and make a guess at how much longer the amp will run, but it would be better if the LED directly indicated remaining battery life. That is the purpose of this circuit.

FET and RFET make a constant current source, keeping the LED at a constant brightness even as the battery voltage drops. You will need to try various values for RFET to get a given current through the FET, since each FET is different. I just touch values into the position with the amp's power on until the LED's brightness is acceptable. Alternatively, you can pick JFETs by testing them for IDSS and jumpering across RLED. A third option is to use a CRD instead of FET and RFET.

ZNR is a reverse-biased zener diode. In this circuit, it indirectly sets the voltage at which the LED will shut off. NiMH cells are spent at about 0.9V. To give some warning between the time the LED turns off and the amp starts sounding bad, let's make the LED turn off when the cells are at 1.0V apiece instead. Let's say you have a 12-cell battery pack and a 5.1V LED. A common zener voltage is 6.8V; if you add that to the LED voltage drop, you get 11.9V, close enough. The batteries are dead at about 10.8V, which should give us enough warning before the amp starts sounding bad. If not, you can swap out the zener for one with a higher voltage drop until you get the behavior you want.

The holes for the zener aren't big enough to accommodate big power zeners. You should only use the DO-35 type, as their legs are thin enough to fit in the holes.

Optional? Yes. Do not jumper.

This is a pad near RLED. You put a CRD (Current Regulating Diode) across RLED to this pad instead of using FET and RFET. A CRD is in fact a FET with a source resistor, trimmed to a specific current value and packaged as a small 2-lead device. They're much more convenient than making your own current source with a FET and resistor, and accordingly they're more expensive. If your time is valuable, they're a good deal.

I recommend the 1N5283-1N5314 series CRDs, which are available in the DO-35 package.

Optional? Yes. Do not jumper.

Resistor Sizes

Except for the output resistors, the resistor pads on the PPA board are only 300 mils apart, which limits the size of the resistors you can use. Standard 1/4W metal film and carbon resistors will fit in the board without a problem.

If you use Vishay Dale CMF series resistors, use the RN55 series. These are specified as 1/8W, but at the temperatures your amp will see, they're actually good for 1/4W.

The exception is the output resistors, of course, but they're big enough for 2 W resistors.

A power supply voltage somewhere in the 10 to 30V range will serve you best with this amp. More voltage will probably hurt more than it helps, and lower voltage will require very careful part choices to make a workable amp. For the full ugly details on how to measure and calculate your way to the ideal power supply voltage level for your situation, see my article Op-Amp Working Voltage Considerations.

If you're going to use batteries, you must use rechargeables; alkalines are not capable of continuously putting out the high current level that this amp requires. There is a companion battery board which includes a charging circuit so you never have to remove the batteries from the amp until they're completely dead. See that section for info on how using the battery board affects the wall supply voltage you need.

If you want a wall supply that's better than almost any commercial offering, take a look at the STEPS. (Bench tests.) While the PPA is less sensitive to power supply noise than many headphone amps, better power quality never hurts.

Whatever power supply you use, it must be an isolated type: i.e. none of the output leads can be electrically connected to the input leads. If it isn't isolated, the virtual ground setup of the amp will often interfere with the grounding setup of other equipment plugged into the amp. Also, if you use a metal case, virtual ground will probably be tied to the case through the pot chassis and/or the input jacks. In order to avoid the problem of tying V- to virtual ground through the case, the DC input jack needs to be isolated from the case. The recommended DC input jack is plastic for this reason.

Stepped attenuators can be better than potentiometers for reasons described in my article, Notes on Audio Attenuators. There is a stepped attenuator built as an ALPS RK27 clone, but I tried it once and I didn't like its performance. If you go with a stepped attenuator, it'll most likely be bigger than an RK27, and you'll have to hand-wire it to the board. Keep this in mind when selecting your amp's enclosure.

The most useful value for a headphone amplifier's volume control is 50 KΩ. If you choose a lower value, the source can have a significantly harder time driving the amp, in which case the source will sound worse. If you choose a higher value, the amp's noise floor will rise, possibly audibly. Some people choose 10 KΩ to lower the noise floor when they know their source is strong enough to drive it, but we haven't found the noise from a 50 KΩ volume control to be a problem. (See the benchmarks.)

Choosing an Output Jack

It's not obvious from studying the PPAv2's circuit that if you're using a metal case, your output jack needs to be an isolated type. A non-isolated jack connects the ground connection to the chassis; since the chassis is probably tied to input ground by way of the pot or the input jacks, this will short out the ground channel. At best, shorting out the ground channel makes it useless, and at worst it will cause it to become unstable.

The easiest way to deal with this is to use a fully isolated jack, like the Neutrik NJ3FP6C or the Switchcraft N112B. (Part numbers for both are given in the parts list.) Another common type has plastic mounting threads but there's a metal contact that goes to ground and is meant to touch the inside of the panel when you mount it. You simply have to slip a plastic washer onto the jack's snout before mounting it to make it fully isolated. The Switchcraft RN112BPC jacks mentioned in the parts list are this way. Keystone has a line of nylon washers that you can use to isolate this sort of jack.

On Output Shorts in PPAv2

The output buffers in PPAv2 do not have output current limiting. That means that it is theoretically possible for an output short to destroy either the output resistors or the output transistors, or both. Since the tip-ring-sleeve plugs used on headphones have a design flaw that shorts the right channel to ground within the jack while pulling the headphone plug out, it's worth considering the consequences of an output short.

The output resistors are the easiest to deal with. We made space for up to 2 W resistors here. To push 2 W through 2.2 Ω, you need about 2 Vrms. That much voltage is definitely excessive with a lot of headphones, but it's only pushing the safety limit a little bit with some. Raising the output resistor to 4.7 Ω pushes the voltage requirement up to about 3 Vrms, a comfortable safety margin. The highest value you should use is 10 Ω — distortion gets too high otherwise — which allows you to get away with putting over 4 Vrms across this resistor.

There's another aspect to the output resistors: their presence means the output is never actually "shorted": there's always at least 2R in series with any output short, where R is the value of R24/34. Henceforth, we will use this value, rather than considering true short circuits.

As for the output transistors, there is no simple answer here: one has to do the engineering.

Let's try to get a handle on what it takes to get into trouble with the MJE243/253. We'll say that the maximum output voltage will be 6 Vrms, which is about 8.5 V, peak. This transistor's safe operating area curve tells us that the current limit is about 2A with that much voltage across it. To force that much current through two 2.2 Ω output resistors, we'd have to develop 8.8 V across them. So indeed, we are barely able to exceed the limits here. It's a bit of a stretch, though, because 6 Vrms is excessively loud in all headphones I'm aware of. For practical purposes, then, it doesn't seem likely that you can kill MJE243/253s with 2.2 Ω output resistors.

Just looking at the front page of the BD139/140 datasheet tells us we're not going to be so lucky with this transistor. It has a 1.5 A nominal maximum, and with our self-imposed voltage limit, the SOA curve tells us that the real current limit is more like 1 A. We only need about 3 Vrms to push that kind of current through 4.4 Ω. But again, this is still louder than anyone should be listening. We're on much less firm ground here, but it still seems reasonable to say that we're not likely to blow up the output transistors.

There's another side issue that isn't terribly relevant, but it's worth examining it to see why because so many people get caught by it: the power supply. A simplistic view is that a wall power supply is likely limited to something on the order of an amp or two. That's within even the pessimistic limit we set for ourselves with the MJE243/253 transistors. The thing is, the rail cap bank has to discharge before that limit matters. The rail cap bank impedance probably isn't even 10 mΩ. You can get stupid-high currents with even low voltages with that little impedance in the way. And if your power supply is a battery pack, it's going to have a continuous current ability of several amps and a flash current ability up in the tens of amps. My advice is, don't depend on power supply current limiting to save you. Add a fuse at the amp's DC input if you wish, but it's the downstream components where you should concentrate your efforts.

In conclusion, there are two practical things you should consider if the lack of output current limiting worries you. First, use MJE243/253 output transistors, or something else equally studly. Second, consider raising the output resistor, possibly as high as 10 Ω.