AM8430 JP-4 ADSR


AM8430 Mockup

Overview The AM8430 module is a replica of the ADSR envelope generators used in the Roland Jupiter 4 analog synthesizer from 1979. It has a very unusual design that was necessary to create voltage controlled envelope timing that could be stored in patch memory. Three voltage controlled CMOS oscillators, one each for Attack, Decay and Release, are in each envelope generator.

The ADSR produces the usual exponential curves but with microscopic steps that are not audible. The design is rather parts and labor intensive, and the 5x CMOS chips were replaced by a single custom IR3R01 chip in the Jupiter 8 and Juno 6/60 from 1981-1982, and then by digitally generated envelope generators from 1984 onward. The IR3R01 looks to be closer to the CEM3310 design than the design in the JP-4. (using CMOS oscillators results in lots of external capacitors).

Back Story In 1978 the only available VCADSR chip was the pioneering Emu Systems SSM2050 that was used in the Prophet 5. This has some minor downsides such as Decay offset at the output and the need to carefully match them for polysynths. Roland were not about to use this chip (nor could SSM supply them in large volumes), so they must have looked at traditional VCA based VCADSR’s which have poor performance (two were used in the Oberheim OB-1 in 1978) and rejected the idea.

The Roland engineers needed a reliable VCADSR that could be used in a polyphonic synth, where the timing needs to be closely matched, and without appreciable output offsets. They took the basic idea of CMOS analog switches controlling the ADSR stages and then added some new innovative ideas.

ADR Tapers

How does it work? The JP-4 ADSR is a unique design which uses high frequency CMOS oscillators and analog switches to create the traditional ADSR envelope generator. They work well, but have some downsides, especially the large number of components needed and the dependency on a stable -15V power rail.

Each envelope generator has three high speed CMOS oscillators with a very narrow pulse width. A 4069 chip (Hex Inverter) is used to create the oscillators with a timing capacitor of 150pF (Attack) and 47pF (Decay and Release). The Attack oscillator is slower with a range of 10KHz to 1MHz and the Decay and Release are faster at xx..

The frequency of the oscillators is varied by a 0 to +5V control voltage that the JP-4 reads from 50k slide potentiometers on the front panel. The timing capacitor of the attack oscillators is 3x larger at 150pF than the decay and release oscillators (47pF).

The A, D and R pots have additional 22k resistors between pin 1 and 3 which makes the taper a reverse log and helps setting values at the lower section of the pot travel. See diagram for the actual taper curve. Roland used the same trick in the JP-8 A, D and R pots as well as on the LFO Delay pot.

ADSR Block Diagram

Each pulse train switches a SPST analog switch (4066) on and off many times per second, which discharges or charges up its own timing resistor and capacitor (C17, C18, C19), which creates the envelope stage voltage curve (A, D and R).

Each pulse takes 470ns to create a new voltage step (the time taken to discharge or charge the A, D, R timing capacitors), which is held on a JFET op amp and a 100nF “hold” capacitor. The pulse train keeps getting another 470ns burst of voltage to either build up the curve in Attack mode or reduce it in Decay and Release modes. The step duration is driven by the CMOS oscillator frequency and these micro steps approximate to a curve.

Even at the shortest attack time there are 3000 steps in a curve and the charging and discharging is exponential. The “steps” cannot be discerned even on an oscilloscope. At 10KHz the step duration is 100us, and at 2.2MHz they are 454ns.

The faster the oscillator the shorter the duration of the envelope stage, a 100KHz frequency creates a 10ms envelope time. Once the attack voltage reaches the voltage set by R105 and R106 (+9. 43V) the comparator Op Amp (IC21) inverts the state of the IC25 flip flop circuit. Using IC24 (triple NOR) the Attack phase is turned off and the Decay stage is switched on using the Sustain level (voltage) as the end point. A similar transitions is made into the Release stage once the gate is off.

The CMOS oscillators are trimmed on the Module Controller PCB to create the correct envelope timings, and to match the timings across the two envelope generators. It is not clear why Roland did not use one set of oscillators driving both envelope generators to reduce component count and ensure the same timings.

Overall Envelope Shape

These single turn cheap trimmers in the JP-4 are very sensitive, and sweep the clock frequency across many kHz. It is really hard to align the timing across the four voices in the Jupiter 4. and a multi turn trimmer is a much better solution.

Attack timing varies from 0.6 ms to 3 s according to Roland but the fastest timing we obtained was around 1.4 ms. The Decay and Release envelopes vary between 3 ms and 10 seconds, faster than the 14 ms minimum quoted by Roland in the user manual. The envelope generator produces 0 to +10V control signals (I measured +9.90V), and an inverted envelope is also available.

AMSynths Design The AM8430 is an exact replica of the original circuits, adjusted for 12V power. The envelope generator uses an external Eurorack gate signal which is boosted to 10V to drive the CMOS circuit. There is also a manual gate push button with a red panel LED. The sliders are Bourns 30 mm travel, with LED’s.

Some early Jupiter 4’s used a matched transistor pair in the CMOS oscillators (2SC1583), even though the layout is for two unmatched transistors. The  use of matched pairs seems to be for older serial numbers, but no mention of them is made in the service notes. The clock stability is dependent on the accuracy of the power rails, so precision rails may help the accuracy and consistency across voices. However for a single ADSR module I have kept with +/-12V rails that are not precision.

Fast Attack 1.5 ms

The Jupiter 4 has a reputation for fast envelopes, which is certainly how they feel as the attack is quick and longer times are right at the top of the potentiometers. On the prototype I tried out log potentiometers (wired upside down to give reverse log) which made for better control, and eliminates the need for the added resistor. However its not easy to source PTL30’s with a log taper, so I have retained the original linear pots with adjusted curve.

The AM8430 has been adjusted for 12V power, whilst retaining the 0 to 10V envelope output voltage. This has been achieved by keeping the original attack profile (aims for 67% of + power rail) which results in the correct profile but maxes out at 8V. An Op Amp amplifies this by 1.25.  In the past I have made the mistake of keeping the sustain level at 10V, which makes for longer (incorrect) attack times. Behringer made the same mistake on the Roland 140 ADSR clone.

It is also important to realise that the sustain voltage from the slider now needs to be +8V. In the JP-4 the post DAC voltage (0 to +5V) for the VCF and VCA sustain is doubled by Op Amps IC28B and 29B. I have added an Op Amp to to amplify the sustain level of +5V from the front panel slider by x1.6 to +8V.

Added Feature The The Key Follow feature of the Jupiter 6 has been added, where the pitch CV can control the timing of the A, D and R stages. Larger pitch CV’s result in faster envelope timing, the Sustain level is not changed. The amount by which the timing changes can be adjusted by a rotary potentiometer and a Range switch selects whether the pitch CV is 0 to +5V or 0 to +10V and therefore a spread of 5 or 10 octaves.

The Key Follow voltages can be set from 120% down to 0%, which enables the amount the pitch change to be actually increased as well as decreased.

Connections There are five 3.5 mm jack sockets on the base of the front panel:

  • External GATE input (Euro Rack standard of +5V)
  • GATE output for daisy chaining modules
  • Pitch input for the Key Follow
  • ADSR envelope output
  • Inverted ADSR envelope output

There is also a manual gate momentary switch for triggering the ADSR, as well as a red LED gate on indicator.

Fast Release 3ms

Module Outcome & Availability A single prototype ADSR PCB was laid out in the Autumn of 2018 and ordered as a prototype in November. Testing was conducted on 15 and 22 December and a couple of changes were needed to get to a production version, along with designing a PCB to hold the pots. A second prototype set of PCB’s was designed in June 2022 to explore the design further, and to use SMD CMOS chips and vertically mounted resistors (like the original). This enabled the module width to be reduced to 10HP.

The design proved successful, and a production version was created with the SMD chips and horizontal 1/8W resistors which are smaller but THD. The second prototype acted as a test bed for getting the envelope timing accurate when using 12V power and achieving the original Roland specifications.

The complex design means it is too big to fit a dual ADSR into less than 20HP, so I have created a single 10HP ADSR module with the additional features mentioned above. Check the webstore for more details and availability.

 

 

 

Copyright AMSynths 2022