Music and Synth DIY

Circuit bending, Schematics, Synth DIY, Through hole

Casio HT3000 – NJM 2090 VCF modifications

First keyboard I ever had was a Casio MT-68, and the first actual synth I ever owned was a Casio CZ-101, so I have a love of old ’80s casios (I’ve got a load of them, including the HT6000, CT410v and MT-400v, VZ1 and of course a couple of MT-68s).

I missed the Casio SD range – the preset MT-600, the “actual synth” HZ-600, and the oddball HT3000, 6000 and mini key HT-700 – when it was in production in the late ’80s, and only found out about them ten years ago or so.

With the exception of the HZ-600, they all have “home keyboard” styling and functions – cheesy ’80s autoaccompaniment, PCM drums, etc – and rather than the filterless Phase Distortion synthesis of the CZ range, they’re DCO voices with a proper analog low-pass filter – the NJM 2090 (two in each keyboard – one for the main voices, one for auto-accompaniment – and eight in the HT6000, one per voice…)

The HT3000

Do they sound like an analog synth with DCOs? Nope. Firstly, although there’s this analog VCF, it’s totally hamstrung. The control voltages in the stock synths won’t let it self-oscillate, and the maximum resonance setting is very timid.

Secondly, the DCOs are four-bit resolution, so they become very steppy on a scope at low frequencies and don’t sound anything like, say, a Juno 60’s DCOs.

The main problem, though, is the resolution of the parameters and the lack of knobs. Programming the synth is, in true ’80s “paint the hallway through the letterbox” style, done with the single data entry wheel. No real-time tweaking here.

This data entry wheel is a voltage divider pot giving 0-5v, which is quantised to five bits by the ADC – so the maximum resolution of the parameters is 0-31.

Still, I love the sound of these very odd keyboards (especially the Ring mod on the HT-6000) and I got to thinking what could be done to mod this thing and make it more useful to me.

NJM 2090 VCF chip


Here’s what I added:

  • Filter envelope with pots for ADSR + LFO to modulate the parameters.
  • Manual Cutoff/Resonance pots
  • Resonance LFO
  • VCF LFO with another LFO to modulate parameters
  • LFO for the data entry wheel control voltage (so an LFO can, for example, switch waveforms while the note is playing, or modulate the DCA release or whatever)
  • MIDI clock sync for all LFOs

If you want to mod your HT-3000, grab a copy of the very detailed service manual – there’s a load of detail on the NJM 2090 VCF including a pinout and details of the control voltage range.

I used the excellent Electric Druid PICs for this mod – the ENVGEN8 envelope generator for the ADSR, the TAPLFO3 for the VCF LFO, four STOMPLFOs for the parameter modulating LFOs, and a MIDICLK for syncing it all up to an incoming MIDI clock with two separate rate dividers.

Control pots board


First problem was that the VCF is the only discrete analog part of the voice. There’s no hardware to take a gate from, and therefore nothing to trigger an envelope – it’s all software generated, along with the DCA.

Because it’s a programmable synth, a gate can be got from the LSI by programming an “organ” envelope into a voice – immediate attack, no decay, full sustain, immediate release. I modded a battery holder into the case, because unless you have the six (I think) huge D cell batteries installed, there’s no backup for the internal programs.

Then, I cut the traces for the cutoff and VCF envelope between the LSI and the op-amp LA3658-2 that mixes them together. The output of the ENVGEN8 is 0-5v – I scaled this to the 0v to 3.5v the LSI produces for its envelope CV, and injected it back into the op-amp, where it’s mixed with a 0.8v-2.5v scaled control voltage for the cutoff frequency. The “gate” for the ENVGEN8 is taken from the CV produced by its internal VCF envelope with my “organ” envelope preset (see the schematics for info).

The trace for the resonance going into pin 9 of the NJM 2090 needs cutting too, and the replacement Resonance CV is injected here. It will quite happily self-oscillate, outside of the CV range Casio set, but again the CV needs scaling (more than about 3v – I think – produces nasty high pitched screaming).

Filter section of service manual

The data entry wheel is, like I previously mentioned, a straightforward 0-5v so its easy to switch between the stock pot and an LFO – this is probably my favourite part of the mod. Set the Parameter Access to DCO Waveform, switch it to LFO, and it steps through the waveforms like an ultra-shite wavetable synth. It’d be great on the HT6000, modulating the ring modulator…hmm.

I included an LFO each for the parameters of the VCF envelope, Resonance and VCF LFO – big thanks to Mark Rundle in the “synth DIY for non-engineers” Facebook group for helping me with the CV mixing circuit when I got stuck.



When I get round to it, I’m going to make a PCB and panel, and iron out a few problems – the MIDI clock is audible in the line out on this first version, it desperately needs some LEDs for feedback on clock rates and LFOs. I’d also include a variable offset for the data entry LFO, as waveform 6 is “noise” and jars as it plays through the different waves, so starting from 7-31 would be better. Most importantly, I’m going to kill the infuriatingly typical Casio circuit that turns the synth off if you don’t press a key within six minutes (“auto power off”, thankfully documented in the service manual).