Tweaks
Building for Tweakability
Virtually every hole in the board is on 0.1 in (2.5 mm) spacing relative to all the related holes. Also, each resistor’s holes are 0.3 in (7.5 mm) apart. This means you can use DIP IC sockets, SIP socket strips and SIP header pin strips in the board. If you did this for everything where it makes sense, you would in effect have a PIMETA breadboard for maximum tweakability. :) Here’s a list of all the places you can do this and ideas for why you might want to do each one:
- R3-R8 block — Use a 300 mil DIP-20 socket or two rows
of socket strips to allow easy swapping of resistors.
- Input and outputs — Solder SIP pins in the holes, and
attach matching female connectors to the ends of the hookup wires to
allow easy disconnection of the panel jacks. (The Molex KK series of
connectors works well for this.) This might be useful for
disassembling your amplifier for servicing without removing the jacks
from the panels — just disconnect the jacks from the board, lift
the board out, and the panel components can stay in place.
- Power pins — Similar to above.
Stacking Buffers
The BUF634 buffer is stackable. This means you can literally stack them on top of each other and solder their pins together to run multiple buffers in parallel. This makes the amp sound better, for a number of reasons:
- It improves the slew rate of the buffers into loads like
capacitive headphone cables, which lowers distortion. See the load vs.
slew rate graph(s) in the buffer’s datasheet. With stacked
buffers, the load is divided among the buffers. Treble response
benefits the most from this.
- It lowers the output impedance of the amplifer. With most headphones, bass improves as you lower output impedance.
You don’t need to stack buffers to get good sound. It’s a “last 10%” tweaking kind of thing.
As far as I can tell, you need to double the number of buffers each time to get a noticeable improvement, and with each additional level of buffers the improvement diminishes. Adding a second buffer to each channel gives a small but noticeable improvement. Doubling that to four per channel is a bit “out there” in terms of bang for the buck. I haven’t tried going beyond 4 per channel yet, but I suspect it would amount to a pretty tiny improvement.
You might think stacking power supply buffers would help. After all, the benefits mentioned above should help the output ground driver as well, since this is where the amp’s output currents return. Listening and bench tests don’t bear this out, however. I’m not certain why this is, but I suspect it’s because the output ground buffer doesn’t have to slew a voltage like the left and right channels’ buffers do.
If you’re using R11, you need to halve its value every time you double the number of buffers in order to keep the same bandwidth setting.
Removing the Buffers
Just as you can add buffers to optimize the amp by spending more money, you can save money by removing the buffers. All three buffers are technically optional. Just jumper pin 3 to pin 6.
If you remove the output buffers, the op-amp chip will have to be strong enough to drive the load all by itself. Even if the chip is capable of a farily high output current, it won’t sound as good as when it’s insulated from the load by a buffer. Also, you’d have to rearrange the R3-R6 values since the multiloop feedback configuration won’t work without a buffer.
Biasing the Op-Amp into Class A
“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.
The recommended method for biasing a PIMETA’s op-amp into class A is called the “JFET cascode”, which are Q1 and Q2 on the PIMETA board. These transistors have to be chosen in a particular way to ensure that the cascode behaves properly; you can read my op-amp biasing article to learn how to find workable pairs yourself, or you can order tested transistor sets with the PIMETA board. The article also explains how the cascode works, and also talks about several other methods for biasing op-amps into class A, most of which are directly supported by the PIMETA board.
I recommend that you wait to bias the op-amp until you’ve got the amp working without it. It’s no more difficult to add it later. If the amp doesn’t work at first, it will be easier to find the problem if you don’t have to chase biasing problems at the same time. For battery-powered amps, you may choose to leave it out because the amp sounds good enough to you without the bias. For wall-powered amps, the extra current draw isn’t going to matter, but some op-amps don’t benefit from biasing, so you might still choose to leave it out.
Adding Crossfeed
This is explained in a separate article.
This space intentionally left blank. :)