Fix: Oberhiem Matrix 1000

[playthatbeat]

Have had a Matrix 1000 on the broken pile for years, and i really miss it, so i decided to dig about and sort it.. The main problem is it does not respond to MIDI - no MIDI input will register on the front panel. It idles well, and can be coaxed into a drone with freezer spray, but won't play ball with incoming midi. So, this page: http://wolzow.mindworks.ee/analog/m1k.htm has a nice overview of the board, and a schem. scan. this post: http://launch.dir.groups.yahoo.com/g...im/message/840 details the issue with the Sharp PC900 optoisolator blowing, so i poked pin 4 on it with the scope (etc.) and can't really see much happening. (pc900 circled in red)

Moving to the MIDI (comms) chip, and looking at pin 6 (marked red in pic) with the scope & starting a midi patch dump from the front panel i see stuff like the attached picture.. all well & good. But, pin 2 (marked blue), the input pin (from pin 4 on the PC900), just never gets close to that while shoving a stream of chords from my Flame 6 in a row into it.. No sign of any nice data at all, just hash.

has to be the PC900 that is borked.. So i ordered a pair on ebay from a dutch guy for like 2 euro, will swap the fuc*er over soon.

The annoying thing about taking the board out is You have to undo all the regulators, but hey, we'll manage it.

I do remember this unit had a sick voice last time i used it - the usual crap - 1 voice sounds different or quieter or not at all when played, so i'll go ahead and work on that as soon as i can get it responding to MIDI in again..

[playthatbeat]

Cool - got the PC900 - it was actually a PC900V, in it went on a socket and i have MIDI in back

now to the bogey voice i suspected i had..

http://www.youtube.com/watch?v=VSBf0J3uQg8

defo voice 4 is borked - sounds wrong/quiet.. and if i move the CEM 3396 the fault follows it.

i noticed if i swap voice 4 with 6, then voice 4 comes back to life, but BRIGHTER then the other voices.. if i then swap them back again, the chip from voice 6 sounds normal in the position - this means i'll have to do a CAL when i get the new chip to match them all up..

Found a used chip from the UK on fleabay for $30 - happy days

more when i get it..

[playthatbeat]

New CEM3396 arrived, in it went, and it's fully working

Now the bad news.. i changed patches, and 2 other voices sound like shit now.. no bottom end/sub osc in voices 1 and 3 now. I suspect they were always like this, and because mostly they sounded OK, i missed it.

SO. no more slim-body 3396 on fleabay, BUT, a shedload of wide-body versions. In fact, for as little as $3 You can get one - BUT, all the cheap ones bar a few are (pictured as) date 8709 on ebay.. lots of diff vendors, but same datecode on them all.. The dead chip i removed was 8852, that would make it december 1988, and it was a slim-body, so iguess march 87 for a spare batch of wide-body chips is plausable, but STILL.. So i ordered some wide-body ones, but with a different date code to the majority on fleabay, just to be different. lol. I ordered 8624 coded, so thats end of march 1986 i think..

I have no idea if the widebody chip is the same as the slim, i guess we will find out soon enough.

I am having to make an adaptor, to allow wide cem3396 to be used in a matrix 1000, as it takes the slimmer, rarer version. So, i ordered some 'turned pin wire wrap SIL sockets' and some stripboard.. i have plenty of widebody sockets, so i'll knock up something using these bits when they all arrive. I thought i could make a tower adaptor, using 3 slim sockets and a wide one, and shorting out one of the slimmer ones.. but fu*k that, i'll make something a bit more reliable..

so yeh, nearly there..

[playthatbeat]

Job done.. chip convertors made.. all good after a CAL..

[playthatbeat]

Got the batch of NOS CEM chips from eBay - seller was hkutsource - i thought it was going to be a fishy experience when the image in the item listing was 'that' date code, and then when i got a msg from them saying the package had been returned to them as un-deliverable, i was SURE i was being dicked in some form.. BUT, they were sound enough to send it out again, without a surcharge, and they arrived in good working order.. Datecode was 3396 - week 33 of 1996. I now have a nice batch in the rack for the next time the m6/1000 takes a sh!t on me..

I see the seller now has a note re: the september/october Post issues, and has changed it's shipping to originate in Hong Kong rather then China.. This is how most of the far eastern vendors have gone the last few months.. It seems stuff from HK is more likely to turn up in one go then stuff from china!

SO, get them in now guys, might as well before they go for good, the price is right @ $6 a pop!

[playthatbeat]

Noticed ysterday a voice has failed again in my m1000.. the usual shit - you can hear that one voice is missing a componant of the patch as it plays. This time i was using a 'doof' tom-tom sound, and every 6th note was only noise, no OSC componant. so, time to get out the screwdriver again

How long was that? how long did the last fix last? (see date of previous post..)

[TheBoyEclectic]

Matrix-1000 Woes

Greetings PlayThatBeat!

Well, I'm running into an issue. I thought that I just had one of the CEM3396 chips going bad, but after running the voice diagnostic and switching a couple of chips, the voice issue didn't change.

Also, I have found that depending on what patch I have selected, I notice one of two problems.
The first issue sounds like one of the two oscillators isn't triggering on one voice (currently voice #1 based on the diagnostic). Voices #2-#6 sound fine.
The second issue on other patches sounds like voice #1, #5, and #6 are almost inaudible.

When I opened the unit up and swapped a few of the CEM3396 chips, it made no difference. The same voices were behaving the same way.

Any thoughts on this? Your insight is greatly appreciated!

Oh, and when I was thinking that my issue was related to the CEM3396 chips, I found the following. I have not done business with this company, but this appears to be the right chip for the Matrix-1000 and it isn't a bad price.
http://www.aliexpress.com/item/CEM33...895826382.html

Thanks!

[playthatbeat]

That Link to Aliexpress looks good! as long as they actually send You that exact format of the chip it will work fine in the M1000 - thats the right shape for the sockets. Shame these didn't appear available when i needed them a while back! I'm going to order a pair of these now, since my M1000 is misbehaving on 1 voice now, and i'm guessing it'll be a slim-body chip that has gone down on me.

Are You sure all of Your chips are working? If you run only 1 chip at a time in socket 1 (or whatever socket works..) , then check all of them using the same test patches - do they all have proper function of:

*square and saw waveforms
*lowpass and highpass filters
*noise generators

You will find a bad chip quickly by just listening. remember some chips might ONLY lose the waveforms, or ONLY the noise generator, while other things function normally.

The totally dead voices sound like they might be related to the 2 types of chips in each voice.. one is the TC4051 datasheet here: TC4051,4052,4053BP/BF/BFT - Toshiba a multiplexer, and the other is a 74hct32 datasheet here: 74hc32 which is a quad OR gate.. Both of these carry out routing and switching to and from the chip, on instructions (down the line) from the CPU and system software. They are the 'knobs and switches' on the synth

If you look at the datasheet for the CEM (datasheet is here: www.synthtech.com/cem/c3396pdf.pdf ) you can see most of the pins are just inputs for control voltages.. You can control all the things the synth on the chip should be able to do using these. You can test these pins to see if any voltages are _actually_ getting as far as the chip when you play. if they ARE, but no sound is coming out of the output pin, then the CEMs have to be the problem. If there are no voltages ending up at the input pins of the CEM, or the state is not changing correctly at the pins as you play, then the 74 and the TC have to be the issue.

on my matrix 6 i had to replace these chips to get the voices to behave. was a bit irritating, but the chips were cheap, and it made things good.

[playthatbeat]

Testing my m1000 now.. defo a voice issue.

To View the Voices as they are played, and find the dodgy voice, i'm doing this:

in "Ext Funct" press 7 then Enter, press + to display "1" Press Enter again & the display shows the voice as it is played..

http://www.youtube.com/watch?v=Rlf6Y2jitaA

My first run showed voice 3 and 6 are not happy. Looking from the front of the m1000, that means the 3rd chip from the right, and the most left chip. (yes, a bit backwards.. right to left increasing in value!) So, i swapped them with their working neibors, and the fault moves with the chips. the fault is somethign to do with the VCA or filter? it makes some noise, but it's wrong, and low level.

OK. time to make a call on weather to build more sockets, or hit Alibaba at the link above..

[playthatbeat]

quick blast on fleabay, and it's got shedloads of widebody chips, from a couple of batches. meh, i have 4 of them here, but only enough bits to make 1 socket adapter.. so, gonna grab a pair of the slim-body chips from alibaba now, while i can. yes, it's more expensive then buying more long 'turned pin wire wrap SIL sockets', but wtf, who cares. Let's see how long they last. Maybe there is a reason the slim-body chips are so scarce - MAYBE THEY ARE SHITE!!

[playthatbeat]

we got dicked..

So, chips from Aliexpress (alibaba.com) arrived today - see pic.. They are NOT as pictured in the original order page (the 'before' pic below) in fact SLIMBODY but are the common WIDEBODY chips (see the screengrab of the product page as it looks today 'after' attached) .. Datecode is 8649, so 3rd last week of 1986.

So, let the warning go out - if You see slimbody chips for sale from china - prepare to be dicked.

So the next thing to do is a step-by-step of making the convertors for the wide to slim.. let's start by ordering some 'Turned Pin Wire Wrap SIL Socket 32 Pin' on ebay - 1 of these is needed for each convertor - and some spripboard for the chassis - get what you can afford, and some 'Turned Pin DIL IC Socket 15.24mm 24 Pin' - 1 per convertor - i got them all from http://spiratronics.com/ in the UK (also on ebay)..

(sigh)

so.. more when all this is here..

[playthatbeat]

OK, let's make 2 adaptors to allow widebody CEM 3396 to fit in a matrix 1000.

Gather Your parts: Enough 'Turned Pin Wire Wrap SIL Socket' - these are the long single chip socket rails - to sit in the chip socket on the matrix 1000 main board, so 2 x 12 hole strips, and any size 24 pin DIL socket - i used what i had for the socket part, then cut the middle away so i was left with 2 single rails of 12 pins, short legs. I actually HAD short-legged SIL rails in the bag, but because the sockets were gold plated, and i'm a hopeless sucker for gold-plated nonsense, i used them instead. This was a mistake as You will see later, as i ended up having to file the short pin rails' edges, and then bend one of the long pin rails after attaching, as it would not sit straight due to the 'stubs' left over from being a DIL socket. ah well. read on, you'll see. But for less trouble, use short-pin SIL rails, don't break a socket like i did.



Then measure out the area you need to seat a chip with strip board



Score the shape on the copper side with a hard knife repeatedly, so you can 'snap' along the mark cleanly (remember: strips run from side to side on the board...)



Once You have 'snapped' away your board pieces, break the strips in the middle as best You can - i have a tool for this, but you can use the hard knife again, a drill bit, anything - just break that copper strip in the middle



Then attach the short-pin rails as shown, pins on the copper side ofc, making them properly aligned so the wide-body chip can sit in the holes (try and remember to attach them 'stumps out' if you use a broken 24pin DIL - see below)



Then add the long pinned rails in the same way, making sure they are the right distance apart so as to meet the socket on the matrix 1000 properly.



If like me, you made your 2 short rails from a 24pin DIL socket, and forgot to attach at least one of them with it's inner wall facing out, so the stumps dont interfere with the long rail fitting in the holes next to it, You may have to file the inside of the short rail, push the long in as tightly as you can, then bend it's long pins to fit after soldering. Yeh, i know, but that's what i get for rushing this!



Then You can mount your finished assembly in the sockets of the Matrix 1000 - here i have replaced voices 4 & 5 with my converters, and the 2 chips i got from AliExpress.



Sitting the converters in the socket is tricky - you should align one side at an angle from above, gently bending the capacitor on the bottom right corner out of the way, rock into the socket gently, then straighten up from above, aligning the other row of pins into the socket, then when you can clearly see all the pins are sitting well, apply uniform firm pressure on each side of the converter - it will 'clunk' slowly into place. once it is in, You will never have to remove it, so do it slowly and get it right. Because the converter stands away from the board on it's long pins, It may *look* like this will touch the case once it's back on, but it does not.

So, my M1000 now has 4 types of chips of different dates and sizes. The original smallbody from December 1988, the smallbody i got on ebay from December 1990, 2 widebody from June 1986 and now 2 widebody from December 1986. This means ONLY ONE of the chips it shipped with still works. To be sure i have a uniform sound, i started the CAL function, and after it i cannot tell the difference between the various voices. The purist nerds in the room insist the slimbody chips have a brighter sounding filter for the same value - as if the chip has it's filter cutoff offset as slightly more open then the widebody chips, but this was inconclusive as the patches showing this had heavy modulation and it was difficult to establish a pattern beyond question. Same for envelope release value - it appeared the widebody chips had a slightly longer decay for the same AENV value, again, this is probably bollocks.

[playthatbeat]

I just noticed that the video i put here: http://www.mrspring.info/vbulletin/s...ll=1#post10132 shows that chips 3 and 6 were not working, but i replaced 4 & 5 in the post above - this is because i moved the chips about to make sure the fault moved with them. No panic.

[playthatbeat]

Yeh so, another fu*king voice dies.. SO, in i go and decided to just go full widebody CEM in this machine.

made 2 more convertors - and there is a slight issue with VOICE 6, as it is right next to a riser for the audio out or something - see the pic - so, had to trim the board real tight to make it fit, BUT it does fit, and off we go with total widebody chips onboard now.

I still have 1 original slimbody chip working, but i decided to remove it and go wide.

PROOF that slimbody chips are bollocks == all the wide chips are still working fine.

in testing, i notice a slight difference in filter quality voice-to-voice, even AFTER a 'CAL'.. not sure if this is the chips themselves, or the voice circuits themselves - who cares, this is analog, and that means some organic movement

[playthatbeat]

This week ( w/e June 10 2017 ) We heard that CURTIS synth chips are being commercially manufactured again, in plain sight, without any attempt at deception. They have started with the CEM 3340 - the VCO chip used in The MemoryMoog, OB-Xa and more importantly, the Voyetra 8. Interesting that this has happened. I suspect the chinese knockoffs caused this to happen. It;s been long suspected that the 'NOS' chips appearing were in fact recently made without licence. So, maybe the apparent rarity of CEM stuff will now subside?

This is good news. Really, it is.