The (real) Jupiter 8 Clone

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: Kent, luketeaford, lisa, Joe.

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

The (real) Jupiter 8 Clone

Post by ACME » Mon Sep 30, 2019 10:04 pm

Yes, that's right: a new Jupiter 8 clone :tu:. And the special thing about it: It's a real JP8. So no polysynth based on the 8 :eek: , but a real copy. Of course, I also have to accept a few restrictions. To prserve the sound character, that's the main goal.

No new idea. I played with synthesizer modules 30 years ago, but unfortunately the availability of components behind the Iron Curtain was almost impossible :bang: .
I started the JP8 project in March 2018, collecting everything possible (and impossible) about the 8 the web has to offer. I have printed out, studied and fully understood all the designs. After that, it was time to search the market for the availability of special components. In fact, most of it still exists. The only exception is the Roland Envelope Generator (IR3R01), which is not available worldwide. I could only catch a single one (for measurement purposes).
Although still available, but unfortunately at astronomical prices, are the colorful buttons. Also, the high-body design of the sliding pots is no longer on the market.

Well. I can replace the knobs and sliders without hesitation. The PanelBoards change only mechanically. The EGs are replaced by AS3310 and / or ENVGEN8. This does not affect the sound. I have all the other ICs.
Another design goal is to make all boards mechanically and electrically compatible with the original. For example, some owners can replace and upgrade the 12-bit Interface Board with a new 14-bit board. All logic ICs are built in as CMOS HCT or LS series. This saves a lot of current and ensures less heat.
The JP8 gets a MIDI interface. DCB is not supported because it is completely outdated.
..
For nights I have transferred the entire schematics in KiCad. What a pain in the ass! Gradually I made the circuit boards. The arrangement of the components is almost identical to the original.

Image

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

The JP8 CPU board

Post by ACME » Mon Sep 30, 2019 10:24 pm

THE CPU BOARD
..is the brain of the 8. It consists of a minimal Z80 circuit. Only one CPU, some address decoder, EPROM, main memory, program memory. In addition, a timer IC for autotuning and storage including signal processing. Clock and reset circuit complete the bord.

What bothered me here was the backup battery for the CMOS program memory :woah: . In the prototype it is still installed, later I rid off this shit and replaced the expensive 4-CMOS RAM IC against a modern FRAM.

Here is the prototype of the CPU:


Image


And here the current CPU board without the whole battery nonsense ;-):

Image

The new memory is also four times larger, so it can be divided into four memory banks.

Regards
Markus

User avatar
cackland
Super Deluxe Wiggler
Posts: 2199
Joined: Thu Dec 28, 2017 5:42 pm
Location: Los Angeles, California

Post by cackland » Mon Sep 30, 2019 10:37 pm

Wow, this is great Markus. Following this :)

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1575
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Post by devinw1 » Mon Sep 30, 2019 11:21 pm

Holy shite! Nice job man. Following!

User avatar
ultravox
Veteran Wiggler
Posts: 745
Joined: Mon Oct 21, 2013 5:45 pm
Location: USA, Singapore, China, Hong Kong, Japan

Post by ultravox » Tue Oct 01, 2019 5:48 am

Wow - Amazing! Following!
This means nothing to me... Oh, Vienna!

User avatar
mothertongue
Common Wiggler
Posts: 145
Joined: Fri Jan 31, 2014 3:58 pm
Location: Poland

Post by mothertongue » Tue Oct 01, 2019 5:54 am

Awesome, I'm very interested in this project!

User avatar
sduck
experimental use of gravity
Posts: 13865
Joined: Wed Feb 18, 2009 6:03 pm
Location: Vortepexaion, TN, USA

Post by sduck » Tue Oct 01, 2019 12:40 pm

Take a look here - viewtopic.php?t=209322 - I'd be interested in what differences there are.
flickr cloud of sound touyube NOT A MODERATOR ANYMORE

User avatar
rosch
roadkill
Posts: 802
Joined: Thu Dec 23, 2010 12:44 pm
Location: doucheland

Post by rosch » Tue Oct 01, 2019 12:51 pm

Following, looks awesome. The future looks bright :D

User avatar
xonetacular
Super Deluxe Wiggler
Posts: 2579
Joined: Mon May 21, 2012 10:35 pm
Location: Boynton Beach, FL, USA

Post by xonetacular » Tue Oct 01, 2019 1:04 pm

um, cool

JanneI
Wiggling with Experience
Posts: 371
Joined: Mon Oct 19, 2015 3:28 am

Post by JanneI » Tue Oct 01, 2019 1:07 pm

Cool! If I'm correct, this is the third Jupiter-8 project here at muffs. I sincerely hope one of these will actually finish. :sb:

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1575
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Post by devinw1 » Tue Oct 01, 2019 3:25 pm

JanneI wrote:Cool! If I'm correct, this is the third Jupiter-8 project here at muffs. I sincerely hope one of these will actually finish. :sb:
This is at least 3. I think it might be 4. Exciting times (assuming 1 or more every gets finished). :hihi: :sb:

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

Post by ACME » Tue Oct 01, 2019 6:31 pm

THE INTERFACE BOARD.
It is getting interesting now ;-)
This board is located above the CPU board and has identical dimensions. Functionally, it was tasked with the generation of the keyboard control voltages and the processing of all signals that should ultimately reach the CPU.

1.) the KCV voltages.
On the board is a 14-bit EPROM-linearized high-precision DAC. From an extremely stable external reference voltage, the control voltages for the 16 oscillators on the ModuleBoard will be generated. The Portamanto generators are also included in the feedback loops of the KCV Sample & Hold circuit. I regret to say, that the portamento circuit uses 4x expenxive IR3109. What a waste at this point. Maybe I will alter this. I have revised the reference voltage generation somewhat. Dont`t want just to copy silly ;-)

2.) ADC.
The 14-bit DAC also generates the comparator voltages for the A/D conversion of all storable analog control voltages. He does this together with the CPU according to the principle of successive approximation. For this he finally creates a staircase voltage between 0 and 5V, controlled by software. For the reconversion of the digital values ​​in control voltages this DAC is not responsible!! Another DAC on the ModCon does the job. Sonically, the 12 of the 14-bit variant can not differ for this reason. The 8-bit DAC on the ModCon does not care how the control voltages were originally digitized.

Furthermore, the card takes over the provision of all signals for the CPU. So almost everything that comes from the controller, as well as the signals from the keyboard via the IF card to the CPU.
I also placed the MIDI interface on the IF card.
------------
Here is a picture of the prototype

Image


The reference voltage generation in the middle of the board is already realized in SMT.
In the current version I went to SMT. Roland has stated in the service manual that the OPAs in S&H circuit are selected types. But nowadays you get the right OPA for every purpose, so I decided against the selection and chose which with low offset and extremely low leakage current. The film capacitors for the portamento must be tolerated very tightly (2%). There are also some in SMD style.

Image

Since the original 14-bit DAC (or compatible) has limited availability, I've already replaced him with a state-of-the-art 14-bit high-precision DAC.
The new Interface works great. The board is thus fully compatible with the original.

Regards
Markus
Last edited by ACME on Tue Oct 01, 2019 10:02 pm, edited 1 time in total.

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1575
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Post by devinw1 » Tue Oct 01, 2019 8:49 pm

Amazing! Markus, question: have you given thought to the mechanics (enclosure/keybed) or is this meant to be a rack/tabletop unit?

I'm sure there's already a next post where you've got it all done though.. :D

Again, fabulous work. You are an A+ reverse engineer!

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

PANEL BOARD A

Post by ACME » Tue Oct 01, 2019 9:46 pm

PANEL BOARD A

nothing special...
Provides for connection possibilities of the whole potentiometers, buttons and the LED display. Contains the multiplexers for analog and digital control elements. Transfers most of it to the interface. It also generates the arpeggiator clock and distributes the two LFOs (coming from ModCon) depending on WHOLE, DUAL, SPLIT to the corresponding module controllers.

Image


I redesigned the digital multiplexer a bit. Instead of the 4x6 channels used in the original, my 3x8 :hihi: . Otherwise it would be boring :hmm: ...

Markus
Last edited by ACME on Tue Oct 01, 2019 10:12 pm, edited 1 time in total.

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

Post by ACME » Tue Oct 01, 2019 10:01 pm

THE MODULE CONTROLLER (MODCON)

With the MODCON, Roland has "outsourced" the commonly used components for the actual sound generation to a separate board. One MODCON controls two module boards, i.e. 4 of the 8 voices.
The MOCDON provides the four additional voltages for the modules (-+13V, +10V, -5V). The -+13V are temperature compensated to keep the tuning stable.
But the main task is the generation of all control voltages and signals, except the KCV (which already does the interface).
There is a SRAM installed on the board, which is fed with data by the CPU at every program or parameter change to always contain the parameter set for the sound just created. The MODCON then works completely autonomously. It reads the RAM continuously. The analog signals are generated in a board's own discrete 8 bit R2R DAC, routed to Demux and then sent to the S&H. From there most signals go to the modules.

In addition, the LFO and Waveshaper are on the board. A noise generator with AGC stage (in the second version of the ModCon) is always noiseing. With a S&H triggered by the LFO, the LFO random signal is also generated.

Here my prototype. It corresponds to the original.

Image

On the board work 7x BA662 which, apart from the noise generator, do not contribute anything to the sound. Besides the original BA662 (SMD version) I also use clones. Since only control voltages are modulated here, I replaced four BA662 with two 13700 on the later ModCon. Does not make any difference at this point. Instead of selecting 082s again, I chose those with better specs in SMD. Later I chose the foil capacitors in the S&H as SMD. It builds better for me and 1206 is still detectable without a magnifier lens ;-)

The 10V and -5V power supply I also redesigned/improved a bit. I just felt like it... And so that I don't always have to buy two different types of OPAs, all OPAs migrated to SMD at the same time. 462 components are installed on the PCB, 83 SMD of them. It is still compatible to the original board.
Here the current version with the above mentioned changes:

Image


Regards
Markus
Last edited by ACME on Tue Oct 01, 2019 10:16 pm, edited 1 time in total.

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

Post by ACME » Tue Oct 01, 2019 10:09 pm

devinw1 wrote:Amazing! Markus, question: have you given thought to the mechanics (enclosure/keybed) or is this meant to be a rack/tabletop unit?

I'm sure there's already a next post where you've got it all done though.. :D

Again, fabulous work. You are an A+ reverse engineer!
Hi,
first there will be a keyboard version. Later maybe a rack version. I don't have a enclosure yet, but I will build it according to the original. I use a Fatar TP9/S 61.

Markus

User avatar
latigid on
Ultra Wiggler
Posts: 952
Joined: Fri Oct 25, 2013 11:05 am

Post by latigid on » Wed Oct 02, 2019 1:54 am

Epic! Best of luck with this amazing project!

User avatar
diablojoy
Super Deluxe Wiggler
Posts: 1618
Joined: Fri Jan 22, 2010 9:03 pm
Location: melbourne

Post by diablojoy » Wed Oct 02, 2019 1:55 am

haha this one might actually make me broke
I do have a broken C8p that might be nice to hack for it though
It probably wont happen today but if it does it definitely wont go smoothly.

User avatar
Arnoid
Wiggling with Experience
Posts: 284
Joined: Mon Jan 25, 2010 4:32 am
Location: Belgium

Post by Arnoid » Wed Oct 02, 2019 2:08 am

Wow :eek: Big project though

Maybe this can help, alot of info and panel decals for the enclosure:
https://jup8restoration.wordpress.com/

But only for personal use :)

User avatar
tjs
Common Wiggler
Posts: 55
Joined: Sat May 30, 2009 7:17 pm
Location: Berlin

Post by tjs » Wed Oct 02, 2019 4:24 am

:eek:

Ich fange schon mal an zu sparen...

User avatar
ACME
Learning to Wiggle
Posts: 40
Joined: Wed Jan 17, 2018 8:39 am
Location: Germany

MODULE BOARD

Post by ACME » Mon Oct 07, 2019 11:57 am

The MODULE BOARD

ModuleBoard is Rand-spech for voiceboard. The Moule is effectively the core, heart and soul of the Jupi and represents all its charm. It contains two voices. Each voice has two oscillators.

Each oscillator is followed by a waveshaper.The shaper generates Ramp-Up, SAW, Square, Pulse and something in sinusoidal style. The noise comes from ModCon.

I didn't ask myself the following question during the replica: should I change something essential about it? Especially here every change in the audio path inevitably affects the character and the sound. Indeat. Therefore, I first made an exact copy.

Here's a photo of the prototype.

Image


In the filters work IR3109 and original BA662N. This need to remain if you want to rebuild a JP8. For comparison I built a multimode filter with 13700 OTA (Jove80). Technically the 3109 is "only" a 4x OTA, but only technical. The 13700 filter also sounds really cool, but not like the Roland filter. So it stays with the original. In the resonance section there is also a Roland chip (662), just like in the HP filter. A 13700 or clone could also manage this task, but it's not the same.

The only issue were the IR3R01s mentioned above, which are no longer available. So I decided to use the CEM Clone AS3310 first, because it also works analog. Built on a PlugIn board, it can be plugged in directly instead of the R01. But the AS have the huge problem that the envelope-times are strongly temperature dependent. Although the data sheet on temperature compensation is weak, I don't know what exactly should be compensated. So a cold 3310 produces 10s attack, the desired 6s at operating temperature after 20 minutes. That's not an option, at least not by today's standards. But the JP8 Service Manual speaks of the fact that the adjustment of the original 8 can only take place after a 30 minute warm-up phase.


Here a variant with AS3310:

Image


A module board consumes massive BA662 (22 pcs) . My optimization approach limits this number to 8. The 13700 can also process the entire modulation of the control voltages. Everything that goes through music stays as it is.


The board has 856 components, almost 200 of them in SMT for the envelopes. All OPA (except the 080s) are also in SMD. The 3310 can't do anything with the A, D and R control voltages provided by the ModCon. They all have to be converted to 0 to -260mV. In addition, there is the mimic for the EnvKeyFollow... Why the effort? The envelopes are almost identical to those of the R01 when warm up!

At the same time I worked on a version with digital ENV generators. Of course, they run stable from the start and require less external circuitry. They are just digital. The maximum attack time can be limited to 5 seconds (or 6) by a potentiometer. This is only important if you want the board to play together with others. It's not crazy to stick with the possible 10 seconds.

Image

The Modules and ModCons are fully compatible with the original Jupiter.
From a purely technical point of view, it is possible to control Modules and ModCon by a microcontroller via Midi only. It's up to you to decide whether you need the autotune or not.
I can adjust the VCOs so exactly that I even could play without AT. Sounds a bit more "analog".
Last edited by ACME on Mon Oct 07, 2019 4:31 pm, edited 2 times in total.

User avatar
AonFLuX
Common Wiggler
Posts: 175
Joined: Thu Aug 13, 2015 4:19 pm
Location: Sweden

Post by AonFLuX » Mon Oct 07, 2019 2:22 pm

Wow, just wow! This will be very nice to follow. Great work!

User avatar
AlanP
Ultra Wiggler
Posts: 969
Joined: Fri Mar 14, 2014 8:37 pm
Location: New Zealand

Post by AlanP » Mon Oct 07, 2019 4:54 pm

Wile E. would be proud, ACME.

More seriously, that is a ton of work (and parts) you've invested in this!

User avatar
pixelforest
Common Wiggler
Posts: 135
Joined: Mon Jan 29, 2018 9:15 pm

Post by pixelforest » Mon Oct 07, 2019 9:53 pm

Very cool, would you consider open sourcing the cad files so that we could lay this out as a primarily SMT project? I would imagine we could collapse this onto fewer and smaller PCBs by switching the resistors to SMT, digital sections to SMT resistors and caps, etc all without losing the components that contribute to sound and behavior by nature of being original style. I'd be happy to lead the work on that and doing test builds of the SMT design.

JanneI
Wiggling with Experience
Posts: 371
Joined: Mon Oct 19, 2015 3:28 am

Post by JanneI » Tue Oct 08, 2019 1:26 am

Open Source Jupiter-8. This would probably have a huge following, both in hardware versions and firmware add-ons. IE. if someone could incorporate MPE/PolyAT for it, that would be huge! (Loving my Deckard's Dream&Kijimi).

I'm not sure if any skilled DIYer is willing to release all his/her hard work for free, but it would sure be interesting for the whole diy community and also for the commercial side.

Post Reply

Return to “Music Tech DIY”