Tweaks

"Class A" refers to configuring an amplifier so that its output devices remain turned on all the time. This reduces thermal variation and eliminates crossover distortion, which makes the amp sound better. You can bias the op-amps in the PPA by adding the cascode JFETs and the source resistor.

Before I get into the PPA-specific details, you should read my article Biasing Op-Amps into Class A. Q1 and Q2 in the article are the same as Q1 and Q2 in the PPA, and Rs corresponds to R9 in the PPA.

To set the bias point once the parts are installed, you simply power the amp up and measure the voltage drop across R10. If R10 is 1 KΩ, then each volt of drop across R10 equals one milliamp through the cascode, so it's easy to trim R9 to get a particular op-amp bias current. 1mA is a good starting value, but you should try other values to see if you can hear an improvment in the sound at different bias levels.

There are three holes near the rear edge of the circuit board. These are for mounting version 1.1 or 1.2 of my modified Linkwitz crossfeed PCBs on the board. You only use two of the holes, the middle one and one outside one, depending on where you want the PCB to be positioned over the amp board. See the previous link for more information.

In cooperation with the op-amp, C7 and a parallel resistance^* in the PPA form a type of "shelving filter", raising the output level of low frequencies: in other words, it boosts the bass. The advantage of this type of filter over other circuits is that it is completely bypassable and it works in conjunction with the op-amp circuit which had to be there anyway so it only adds two components.

^* The "parallel resistance" is either R7 or the bass boost adjustment pot. Which one you use is a personal choice. Just think of the pot as an adjustable alternative to R7. There's no good reason to use both R7 and the pot at the same time. For clarity, I will just say "R7" from here on.

R7 raises the level of the highest bass boost:

C7 changes the point where the bass boost begins:

The larger the value of C7, the lower the frequency where boosting starts.

Notice how finely spaced the lines are on the graph. This is a log-log scale graph, though, so in reality the jumps in corner frequency between two close capacitor values get larger as the capacitor gets larger. Therefore, if you want the bass boost to extend clear up into the low midrange, you must use a line of capacitors with fine-grained steps between values or else suffer from large jumps in the point where the boosting begins. The Panasonic ECQP line will work well here, for instance; they go up to 0.12 µF without requiring lead bending. If you want to choose a different type, you should use caps with polypropylene film-and-foil construction for best sound quality. Metalized polypropylene will also be fine, but I'd only do it if I wanted a really low amount of boost, and so needed a high value for C7.

The value of R4 also has a small effect on the behavior of the bass boost circuit:

I include this graph only to show that you can vary R4 without worrying about what it will do to the bass boost behavior.

It's important to realize that this circuit doesn't roll off at DC, so any DC offset on your amp's output will also be boosted when you engage the bass boost. It is important that you minimize DC offsets in the amp before adding bass-boost. A 20 dB boost (10×) can transform a harmless 10mV DC offset into a headphone-frying 100mV offset!

If you want to explore these issues in more detail before you turn on the soldering iron, I suggest you download a copy of the Micro-Cap 8 demo. Here is a circuit file for you to start with. (This is the circuit file I used when generating the above graphs, so it should be set up properly already.)

You can get some of the information you would get from the simulator without as much work by using the bass boost calculators.

If you're curious as to how this circuit works, look at R7 and C7 as two separate paths through the feedback loop. C7's impedance ranges from infinite at DC down to 0 Ω at high frequencies. Since 0 Ω in parallel with R7 is 0 Ω, at high frequencies the amp gain follows the standard formula, given as [1] below. (Yes, the gain formula is different in the Jung multiloop configuration, but this approximation suffices for our purposes here.) At low frequencies, C7's impedance is very high, so R7's resistance dominates the total impedance; it simply adds to R4's resistance, giving formula [2]. Between these two levels where the impedance of C7 is near R7's resistance, you have the slope you see in the graphs above. Formula [3] is derived from these two, telling you the amount of bass boost relative to the amp's normal gain.

	(R4 / R3) + 1                 [1]
	((R4 + R7) / R3) + 1          [2]
	(R7 / (R3 + R4)) + 1          [3]

The standard Q3 configuration works well, but there's some room for tweaking here.

As I mentioned elsewhere, you can get up to about 30dB of isolation without really thinking about it. You can get better isolation by picking your Q3s such that the current draw of the circuitry "below" them is a high percentage of their IDSS. Because the IDSS of JFETs varies so wildly, you must measure your JFETs' IDSSes and use similar JFETs in all of the positions to get the best performance.

There are two consequences of doing this tweak. One, it makes rolling op-amps more difficult: you may have to change the Q3s at the same time to make the new op-amp work. Two, the voltage drop across a JFET increases the closer you get to its IDSS. This could be a problem for you if it drops too much voltage for your application, or it could be a way for you to use low-voltage op-amps while still using a high supply voltage for the buffers.

You may have noticed that you can jumper across all of the Q3 positions to get a single set of power rails for everything. It's my opinion that if you're trying so hard to save money that the cost of a few JFETs matters, you would probably be better off building an amp with an inherently less expensive design like the PIMETA.

This space intentionally left blank. :)