Almost finished PCB. Only needs a front panel.
picture above shows the mostly populated PCB of the filter I
built. Several parts on the board were replaced multiple times as I
debugged the circuit and made small adjustments to resistor values. I
ran into several problems during the construction. Most of them were
caused by bad solder points on the tiny transistor pairs (SOT-363). The
transistors used were DMMT3904W and DMMT3906W from diodes.com. These
are not reccomended for beginners. I have never had bad solder joints
working with larger parts. The difficulty is not so much in the pin
spacing (0.65mm), but in the size of the complete part. I actually had
to go out and buy a magnifying glass to read the orientation marks on
the BJT pairs. There was some talk on the Synth DIY list about
making 8pin DIP circuit boards to hold a few small surface mount BJT
pairs. I did a small layout for the discrete OTA structure used in this
filter. The OTA plus a SMD darilington buffer can easily fit on an 8pin
DIP board. I checked pricing of boards this size at Futurlec and they
would cost about $0.50 in quantities of about 20. This would be a good
approach for someone who is not comfortable soldering the small parts.
Another, easier option would be to just use hand matched BC550/560s
like the original did. I used the diodes.com pairs mostly out of
curiosity. I intended to take some measurements before building, but
they were just to small to work with unless I had made a circuit board
for this purpose.
The other SMD part (SOIC-14) is a HFA3046. This is a modern replacement
for the CA3046. It is speced for high frequency and it does cost a few
dolloars. I measured the matching on two of them a while back, but lost
my paper that I wrote the results on. I cannot say exactly but if I
remember correctly, they are comparable to the CA3046 and maybe
slightly better. I have used the HFA3046 here because I had a few
samples left over from some previous experiments.
The schematic diagram will be posted once I have the PCB mounted in a
front panel. This might be a month or so because I would like to have
panels made for several modules at once. For now, I have posted a basic
analysis of the filter using linear approximations. Eventually I plan
to write more about the nonlinear characteristics of the filter.