MIOS Project - Oberheim - M1000 XPANDER {PCB traces uploaded}

Design Concepts
1

…turning a Matrix M1000 into a monster Xpander ? :wink:

(pcb is 220x63mm)

Heres a little project for you,

its been sat at the back of my mind for a while, that it should be entirely possible to make a oberhiem matrix xpander / m12 type synth from the venerable m1000 rack units, by realising the full potential of what should be possible with the humble little cem3396 chips they are based on.

It seems abit of a shame that the chips are capable of fast, realtime control changes; yet the m1000 processor is so slow ..it just cant cope & the filter cascade should also be hijackable in the manner seen in the matrix 12 / xpander synths (for massive filter choices)… however they also lacked some niceitys, like -proper- resonance control, continuiously variable stereo pan for voices (preset pan positions) …and having much faster envelopes would bring the chips to life for percissive sounds.

this project intends to allow the user to get the best out of the humble chips, and allow them to realise their full potential - and with a little luck, the project should be a considerable improvement on the xpander synths too :wink:

xpander filter modes:

    * one-, two-, three- and four-pole low pass

    * one-, two and three-pole high pass

    * two- and four-pole band pass

    * two-pole notch

    * three-pole phase shift

    * two- and three-pole high pass plus one-pole low pass

    * two-pole notch plus one-pole low pass

    * three-pole phase shift plus one-pole low pass

Im quite happy to do the hardware work on this, however i dont have the time to fully nail down customising mios to work with this just yet. So, if folks want to see this sooner, rather than later ..someone/s will have to kindly volinteer some coding skill to get this off the ground.

Heres where im upto now with this:

i have combined (most) of the features of the xpander voicing with the cem3396 chip from the m1000, and added some initial enhancements.

* Due to the need for an external VCA to allow the filter mods to function, this also allowed for fully featured CV pan control to be implimented.

* Also, instead of only having the choice of noise, instead of osc b ..the ability to connect any external audio source is provided.

* dual vca added for level control & source mixing between external signals & noise at the same time.

* and obviously the abilty to use all of the xpanders filter modes is now added to the cem3396.

BETA schematic & board traces in full here:

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_SCHEMATIC.pdf

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_TOP.pdf

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_BTM.pdf

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_LAB.pdf

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_DRL.pdf

http://homepage.ntlworld.com/nausicaa.valley/ucapps/M1000_XPAND/XP_M1000_PCB_%231BETA_BOM.TXT

other resources here:

m1000 schematics: http://wolzow.mindworks.ee/analog/files/OberheimMatrix1000Schematics.pdf

(hell to read)

xpander schematics: http://manuals.fdiskc.com/tree/Oberheim/Oberheim%20Xpander%20Service%20Manual.pdf

(conversely, a fascinating read !)

cem3396: http://www.synthtech.com/cem/c3396pdf.pdf

Additional notes:

* vcas may well work fine with less opamp buffering, but playing it on the safeside for now.

* design untested as of this time, some further modifycations maybe needed for perfect fuctioning. also some of the attenuator values for controls & signals may need tweeking for expected performance.

* control system consists of analog CV control for most elements; with filter modes & ext_in/noise switched in binary.

* oscillator frequencys & sync modes controlled by pulse trains from cpu. (ancillary logic controlling Sync modes omitted, as i would expect the cpu to be capable of generating correct signals).

* depending on the practicalitys of interfacing the mios system to the hardware, a proper revision could be made with either one cpu per a chip, or comprimising to a multiplexed scanning system.

* also, it shouldnt be too much hassle (programmers permitting..) to impliment the tuning proceedures for callibrating the oscillators, filters & vcas.. as this is fairly well documented in the xpander lit :slight_smile:

* if a modular approach is taken, should be able to chain as many of these togeather as you want, for silly polyphony/multitimberality :wink: :slight_smile:

..any takers for this project ?

2

wha.. no responses ? :slight_smile:

either im deeply unpopular, or everyones farked off on holiday :slight_smile:

..eh well, will be leaving this up as a curiousity & please drop a note if anyones interested in this.

ta.

3

Sounds freakin’ awesome.  The Matrix 6/1000 are sweet sounding instruments which suffer from slow envelopes and a piss-poor user interface.

I’m enthusiastic to see where this can be taken.  But I’m not volunteering, because I’ve got so many other unfinished projects!

4

Sweet deal!

The thing is that the old 405x chips produce plopping sounds when switched. They are not ideal if any realtime switching is going on. It’s been observed by Marc Bareille that the DG40x chips are more silent. He made this board: http://m.bareille.free.fr/modular1/vcf_multimode/extrapole.htm which is also based on the xpander filter. I would suggest a component change if you want to go “analog Wavestation” with this, but the DGs cost a fair bit.

Earlier on there was a thread about using CEM filters in Design Concepts where I went through some of the intricate bits involved in the VCF CV in order to help a poster who was looking at making a vcf board. Earlier I was experimenting with the CEM3379 thinking that it too could be xpandified. Turns out it can, but it doesn’t have this constant volume/bandwidth trickery that the CEM3372 has, so the VCF output drops a lot with increasing resonance. I might be useful to try out the CEM3396 for this as well before getting high hopes, but the datasheet for the 3396 mentions this and it looks more similar to the 3372 in topology than the 3379 so with some luck it this feasible with a good/useful result.

Now, it’s too bad I don’t have a Matrix 1000 in some closet :slight_smile: There’s only some 3396 chips in a drawer and my wish to blow some life into a potential MB-3396 along the lines of the old user project floating around but based on the new core.

5

yup, the 3396 meantions the same filter behavoir as the xpanders 3372, both of them maintain constant loudness for resonance.

so all should be good :slight_smile:

as for the mode selector chips plopping in realtime use, trying to attain that maybe like trying to run before you can walk :wink: ..it might be a nice extra, but not as useful as getting access to multimode filtering in the first place. also, where and when the rest of this is made to work, theres no reason that 16 sil pins couldnt be dropped in place of the pcb sockets for those chips & various experimental boards hooked on to try different approaches to making that possible. then from there a rev 2 board could be made with the extra feature. but i expect most people would be happy enough with non realtime multimode filter switching :slight_smile:

as for the chips, theres still afew spares floating around & although the various second hand outlets have been strangely quiet on the oberheim synths lately, there are an awful lot of matrix 1000 synths out there.

6

No worries. You’re correct of course: it’s easy to add an alternative solution to replace the 4051/53 combo. The DG408 ferinstance is of course not pin compatible with the 4051 as that would make life less interesting. When I built my old multimode I used an ADC, pot and jack combo for the mode select :slight_smile: It was awsomeness in a very small box  8) but might not be for everyone.

I suspect that RX27 needs to be changed as the signal at cell4/Cd is probably 70 times weaker than that of Ca-Cc. This can be inferred via pouring over and comparing with the 3379 PDF. This is not a problem per se for the Xpander/Matrix 12 since those use the externally available VCF out which is larger in amplitude. Anywho, if need be that’s a real quick change which doesn’t affect the PCB.

Now, I need to start looking at the 4 odd CEM 3396s as well as the copy of the Cheetah MS-6 service manual I got with a couple of them courtesy of Ebay. Damnit, this fall/autumn will be busy if it starts depleting my check later stash  :o

As for nice extras: if you add the frosting on top you could end up with realtime filter mode changes as well as continuously variable waveshapes thanks to PWM and modulating the waveform slope. The M1000 could only do modulation via LFOs as it was never to big on NRPNs or controllers. Almost drool-worthy and very analog sounding.

7

Hi again Artesia! I’ve been looking into this a bit more but my coding skills are rusty to put it nicely. However, I tried to fill in the blanks on the CEM by looking at the schematics of the previous art. Next up is to forge ahead with block diagrams and then pencil in some solutions.

How to get to an MB3396:

  • Voice card hardware (your very nice board).
  • Rudimentary timing system with rough tuning accuracy, can be done with the Core i/o or PWM.
  • Multiplexed control using data distribution and S/H circuitry. The ghetto method (4051s plus caps) can be employed if we can guarantee fast updates.
  • Improved timing, possibly with range switching of the waveform converter. See below.
  • Improved S&H as in the Xpander service manual (small and large steps, smoothing etc).
  • Waveform trimming. See below.
  • Sync mode logic lifted from the Oberheim schemos.
  • Other autotuning for the DCOs (is this needed?) and the VCF frequency/tracking etc.

The lite/lab version of the software needs to grow as the above goals are reached.

3396 timing and range:

There are quite a few things to realize around the 3396 chips for this to be viable. It was observed in the thread http://www.midibox.org/forum/index.php/topic,7466.0.html that the DCOs are best driven using dedicated timer chips (8253/8254) or with a beefier MCU. Also, the data-sheet gives away that active correction is needed for the waveform converter in order to reach more than 6 octaves of range.

I was looking at the Cheetah MS6 schematics available at: http://www.butoba.net/homepage/MS6-schem.pdf as well as the Matrix-6/6r schematic at: http://www.flickr.com/photos/matrixsynth/sets/72157607736211309/ in order to see how that may be done. It turns out that I can’t really see the range trick implemented since not all ICs can be identified in any event. However, the reasoning behind this is given in the CEM3396 PDF pp.7-8 and a solution lies in the last barely legible page. It seems that the other schematics source at Oberheim Heaven contains corrupt files. Better Matrix-6 scans would be helpful.

Waveform trimming:

Also, there is talk about using the processor and successive approximation to trim for the correct waveshape in the CEM PDF. The processor can measure the triangle wave output using an SR flip-flop to detect glitches. I see some circuitry for this in the last page of the data-sheet as well as the circuitry for the [tt]PULSE[/tt] signal visible in the Matrix-6 schematics.

I am going to try to study these schematics in order to decipher which chips have been used for some reverse engineering. I hope someone else can also gleam something from this. Your board can be tapped for the CEM3396 output and the waveshape circuitry can be placed off-board. Range switching seems trickier to place but can be omitted at first in any case.

Then there’s the software :slight_smile:

Suppose that a slave PIC has to update the various CVs at the 4051s if we copy that method. This means that the MIOS node (Core) processes MIDI, takes care of GUI-related tasks, patch mgmt etc and then updates various registers in the slave. Also, someone (the MIOS node?) needs to update the timer if an external timer IC is used. Or we wait for the new Core32 to do all of this. I smell periodic tasks so this won’t be a walk in the park.

Moving on to loftier software goals: It would be really nice to be able to read/convert Matrix-6/6r/1000 as well as MS6r patches whilst observing parameter differences such as lin/log envelopes plus mono/bi/multitimbral settings. This is totally over the top and best done using a proper computer.

There it is… the road ahead to the old, much talked about MB3396 in some form. This is usually how far it gets it seems. It needs takers and doers. I can at least try to fill in the gaps for the 3396 chip (any help is welcome). Over to our OP Artesia.

8

Sounds like a fun project!  Alas, I’m neck deep in other projects right now, and if I don’t ship them soon and get some cash flow, I won’t make rent.  I don’t think my wife would like that…  :o

I had already started working with the 3396 chips, so I’ll join in with you when I’m done with this other stuff…

:slight_smile:

9

Hi,

I dreamed of this synth years ago…

Now … how about the availability ofthe CEM3396

Without these not much will be possible isn’t it ?

(damned)

Joris

10

This is a great topic. There seem to a few threads in various boards about this chip, but it seems like interest never goes past a certain point (app note schematic, smudged ancient xeroxes with lost wisdom.). If Matrix 1000s were worth as much as a M12 I’d bet things would be more active. I guess no one thinks they can outsmart Tom Oberheim on getting the most out of this very obsolete chip. :slight_smile:

11

Well, well, looks like this thread just made The Matrixsynth blog…

12

UP!

Dear Midiboxer,

Is this project still in developement ?

I bought a matrix 1000 1 year ago and control it with my BCR2000 ( a behringer controller)

Unfortunetly the Matrix 1000 Cpu is very F—ink slow.

Any live control are impossible.

That 's why i ask if this project still life ?

Best regard,

Xarolium