Hookup Wiring
You may have noticed that the wire pads are not placed where they
might be closest to the panel where each component is most likely to be
placed. For example, the audio input jacks are usually on the back panel,
yet the input pads are near the front of the board. The reason is that
when laying out the components on the board, we favored short traces and
a minimum of layer changes to layouts that would have made for shorter
hookup wires.
We made this tradeoff because hookup wires are flexible. Bad sound due
to a poor PCB means you have to throw out the whole PCB, but bad sound due
to suboptimal hookup wiring is fixable by using a better hookup scheme.
It is possible that these longer hookup wires could act as antennas,
bringing interference into the amplifier. I haven't found any such
problems in the amps I've built on this board. If you do encounter a
problem, the single easiest and most effective fix is to twist related
wires together: the wires going to each set of input or output jacks,
for example. You could also look into things like rerouting each wire's
path from board to panel component.
During the prototype process, I took a pair of "naked" amps (no cases)
to RFI Hell. I live in rural New Mexico, in a county of about 100,000
people. So, every TV, radio and microwave antenna in the whole area
is up on single bluff overlooking the area's major city (pop. 40,000).
I parked my car within sight of 12 towers, most of which had multiple
antennas on them. All of these antennas were close by — the
furthest was probably less than 200 yards away. My car stereo was doing
funny things in the middle of this RF bath; if my house were up there,
I would be worried about having Funny Lookin' Kids. To cut this long
story short: I got no hum, no hiss, no AM radio through the headphones,
just dead silence, even with the volume control cranked all the way
up and high-efficiency headphones plugged in. I drove back and forth
through the area a few times, and heard no change at all through the
headphones. Satisfied, I spent an hour watching the traffic going by
while I listened to some tunes. :)
Configuring the Amplifier Topology
The four feedback resistors — R3/R4 and R5/R6 — act in
pairs to set global and local feedback, respectively. They set the gain
for each amplifier, and they also have an influence on the noise behavior
and stability of the op-amp chip.
Depending on how you populate R3 through R6, you can set up one of
three different amplifier topologies. In increasing order of preference,
they are as follows:
Buffer outside the feedback loop (local feedback only)
— This is like a CMoy amp with an
EL2001 acting as a voltage follower after the voltage amplification
stage. To set this up, leave R4 unpopulated, jumper across R3,
and pick values for R5 and R6 to set your
amplifier's gain. You can also leave the output buffers out and jumper
across their position (pin 2 to 7) which turns the META42 board into a
kind of über CMoy
.
Optimizing the Loops in a Multiloop-Topology Amplifier
The gain of the inner and outer loops are more or less independent.
Changing the inner loop's gain usually won't change the overall system
gain very much at all. The amplifier's "real" gain is configured by
the outer loop. Use the multiloop gain
calculator to be sure, though. There are edge cases where the inner
loop's gain can affect the overall system gain significantly.
The purpose of gain in the inner loop (R5/R6) in a multiloop topology
is to optimize the bandwidth of the op-amp. For voltage-feedback op-amps
(the typical kind), gain and bandwidth have a linear relationship. That
is, increasing gain by a factor of, say, 10 will decrease bandwidth
by a factor of 10. The op-amp's datasheet will have a spec called the
gain-bandwidth product (GBP) which expresses this relationship. If the GBP
is 16 MHz, then with a gain of 100, you will get 160 kHz of bandwidth.
So, how is this useful? Since we're only interested in the audio
bandwidth (5-30,000 Hz, generously speaking), what use is an 80 MHz
op-amp? Partly we're interested in the higher slew-rate a high-bandwidth
op-amp gives us. The downside is that high-speed op-amps are harder
to make stable. Since we don't actually need more than about 30 kHz of
bandwidth, can we have our cake and eat it, too? Yes. If you intentionally
raise the gain high enough that you get bandwidth down into the 100
kHz range, the op-amp retains its high slew rate while reducing its
bandwidth. Lower bandwidth means immunity to ultrasonic noise (i.e. most
electrical interference) and greater phase margin. Both of these translate
into greater stability.
(Why use 100 kHz, and not, say, 30 kHz or even 20 kHz? It's mainly a
"fudge factor" — we don't want to cut it too close to the audio
bandwidth, in case the op-amp has some nonideal characteristics that
cause it to have problems with the bandwidth limited so tightly to the
audio bandwidth. Allowing 2 to 3x the bandwidth actually required is
safe.)
You can use a gain of 80-100 with most op-amp chips if you don't want
to worry about this. This is suitable for op-amps with GBPs as low as
8 MHz, which is the slowest op-amp you should use for a META42.
Once you have the local loop configured, you can pick values for the
global feedback resistors (R3/R4). A global gain of 2 to 15 is typical
for a headphone amplifier; the proper value will depend on your source,
the pot's value, your headphones, and your music listening style. You
can use the gain goal-seeker to
find resistor values that will give a particular gain value, and then
use the multiloop gain calculator
to find the exact gain you can expect for your chosen resistor values.
The outer loop's resistor values should be lower than the inner loop's
values. As a rough rule of thumb, make R3 five to ten times lower than
R5. Also, remember that the lower these resistor values are, the more
current your amp will use. I personally try to make R3 no lower than
about 220 Ω, which means that R4 would be 2 KΩ for a gain of
10; then, R5 would be 1.2 KΩ or higher, making R6 120 KΩ for a
~100× inner loop gain. At the same time, you don't want to go too
high with these values, either, because that increases the chances that
noise will get amplified along with the music signal. Avoid resistances
over 1 MΩ if you can. Not only do larger values risk noise pickup
in the amp, large-value 1% resistors tend to be expensive.