|
AM 8003 Module - Roland SH-3 and 3A Filters
Overview This
module takes the Ring Modulator and Audio Mixer from the Roland SH-5 and grafts them onto the Roland SH-3 and SH-3A voltage controlled low pass filters. But why are these filters so interesting?
The SH-3 monophonic analog
synthesizer dates from the very early days of Roland, it was launched in 1973 along with the preset SH-1000. It is a 4 octave keyboard synthesizer with a single VCO with octave dividing technology (like an organ), rather than a
traditional VCO. I may replicate the VCO later on, as it is quite unusual - providing a form of additive synthesis. The first version of the SH-3 has an "EMS" style diode ladder VCF, and the second version (contrary to popular myth) has
a Moog style 24dB low pass filter. Okay, so on to the reasons to build one:
- The SH-3 VCF is a 4-pole 24dB diode transistor ladder, similar to the EMS VCS3. It also appeared in the SH-1000, and the filters in the SH-5 and System 100 are similar
- but they use transistors as diodes in the ladder.
- The SH-3A VCF that replaced the diode version is a traditional Moog transistor ladder filter with a 24dB response. The design is very close to that used in
the Minimoog and 904A. Once again this is a unique Roland design, only seen again in the Roland SH-2000.
The AM8002 enables three audio signals to be feed through into either filter, along with Ring Modulation of the first two signals. The filter outputs are mixed together and there is independent control of CV voltages for each VCF. The original SH-3 and SH-3A call the CV control of the filter GROWL.
AM8003 Module The original filters are quite simple and basic designs. Roland used a variety of early Op Amps, including the low quality 741 and 1458. They even used a very old 14-pin dual Op Amp - the NEC
UPC33C complete with heat sink. This chip was usually found in early Japanese stereo amplifiers such as the 1974 Trio KT9007.
The SH-3 diode ladder is essential a copy of the EMS diode ladder right down to the FET controlled
resonance circuit. Whilst the ladder buffer and CV control circuits are slightly different, the only other addition is a 5th capacitor (22nF) below the first pair of diodes. This introduces another pole to the filter. I have replicated
the Roland design but used modern Op Amps and used the SH-3A frequency CV circuit to replace the rather weird UPC33C version.
The filters both use Mylar 68nF capacitors in the filter ladders, and I have matched these carefully
within 1%. The diodes are matched to within 2 mV, and the transistors too - I have used the original 2SC1000GR - although modern equivalents can be used. Dual matched transistor pairs are used at the base of the filter ladders, in fact
the LM394 which I bought a lifetime supply from Futerlec at less than 1 UK pound each. I tend to use them in filters rather than the more demanding VCO circuits, where I use SSM or THAT chips for the improved matching. Matching the
diode in the SH-3 filter is important to minimize CV feed through, which can be significant in a diode ladder.
There are two PCB's in the module:
- SH-3 PCB: implements the diode VCF
- SH-3A PCB: implements the transistor ladder VCF
The module has the following front panel controls:
Frequency Cutoff and Resonance for the Diode VCF Keyboard CV level and 2 modulation CV levels for the Diode VCF Frequency Cutoff and Resonance for the Moog VCF Keyboard CV level and 2 modulation CV level's for the Moog VCF
This in just one 3.5 mm jack socket mounted at the base of the panel, this is the mixed audio output.
SH-3 Outcomes I built the
SH-3 diode VCF first (so I could compare it with the SH-5 version). The build went well except for trying to replicate the old UPC33C "preamplifier circuit that controls the Frequency Cutoff. It has a strange feedback loop using a 220K
resistor. I reverted back to the SH-3A design which uses a more typical 1458 Op Amp - this worked perfectly first time - although I used a TL071. The filter trim has a wide range and quickly enabled the front panel Frequency control to
be setup to provide low pass cut off to full resonance.
The filter did not initially self oscillate at first but after some modification it was oscillating in a very smooth and subtle way and the RESONANCE pot provided just the
right feel. I changed the RESONANCE pot to 10K linear and increased the size of the resistor that feeds the FET to 680K. The reason that some trimming is required is that the pinch voltage of the FET varies between individual
transistors (and I have used a 2SK30A-O which has a lower IDSS rather than the correct Y variety), and although the circuit has a Resonance trimmer - some fine tuning is needed. I also wanted to get the right pot feel, as log and
anti-log pots give too fast a transition from low to high resonance values. Perfect!
Voltage Controlled Resonance? Is it a simple matter to add an Op Amp buffer circuit to enable voltage controlled Resonance on the SH-3.
So maybe on the next PCB run I will add this feature.
Outcomes SH-3A To follow....
Front Panel
The front panel is a 3" wide FracRac with Alpha potentiometers and the usual collection of plastic 15mm mixer style push on knobs.
Why do diode filters sound different to Moog filters? Tim Stichcombe has investigated the transfer functions of both diode and transistor ladder filters, and his papers are well worth a read.
Aside from differences in the input and output stages
in terms of circuit quality and frequency response, the different implementations of feedback to create resonance and the different quality of capacitors in the ladder, there is a key reason why the diode ladder sounds different to the
traditional transistor ladder filter.
In a diode filter the 'isolation' between filter stages provided by the transistors is not present in the diode ladder, giving it a different transfer function which in turn means
the pole placement and their subsequent movement with increasing resonance is also quite different from the transistor version. The transfer version has sometimes been described (erroneously) as 3-pole or 18 dB, this is because the
transition from the passband to stop band starts earlier in a diode filter, and lasts longer, so that the 'corner' in the frequency response is less pronounced and smoother.
Manufacturers seem to have misquoted their
filter response as 18dB when its really 24dB, as well as messing around with an extra capacitor and resistor on the base of the ladder to give another pole and a sharper response at cutoff.
|
