inevitable SEM clone testing

SEM testing box

Everyone’s done a SEM-clone, haven’t they? Or a SEM-type of thing. There’s a good reason for that, they do sound really nice.

I’ve been plotting this for about five years or so. I was stymied for a while by the designator-less PCB layout in the service manual, but finally got some boards made just before Christmas. I went through the pain of putting it all together in a PCB box just to test, seems like it works quite nicely.

SEM clone board

Maybe too nicely, I need to work up the motivation to make a panel PCB.

I went for multi-turn potentiometers for the presets, 0.1″ headers for the connectors rather than whatever freaky Molex connectors were on the original because I’m cheap, and kept the component spacing and track layout otherwise as far as possible.

I’m using 2N3906 and 2N3904 for the BJTs, and J112s for the FETs just because they seem to be available. The surface-mount adaptors for the LM301s seem to work alright (as long as you actually solder the legs…), and I kept with 741s. I’ve heard that the reissue uses TL062s, for whatever that’s worth.

I’ve bunged the usual CA3046 sub in for the CA3086. I was intending on using a surface-mount version in a DIP adaptor, but then found that the usual adaptor is too wide to fit in the standard socket. In retrospect I should’ve just actually used a surface-mount footprint on the board.

I stuck with the 723 regulator which was probably daft because they’re obsolete, but then I’m probably only going to make a couple of these. I found some probably-way-too-big silver mica caps for the filter capacitors, and some weird massive looking yellow film capacitors on the filter input, mostly chosen because they might be terrible. Maybe in a good way.

Mistakes that I made and discovered so far include swapping the designators for R73 and R75, a couple of transistors (Q7 and Q16) are flipped around the wrong way, and for some reason it seems like I just gave up when it came to the F connector in the middle of the board.

Let’s look at some wires, the horror behind my quick test setup. My wife thinks I eat them or something.

SEM temporary box wiring horror

I’ve deviated from my usual tactic of employing an Oakley PSU run off a Yamaha PA-20 (which works really well), and instead trying out a switching power supply from China off eBay, fronted up by an LM317/LM337 in a vague attempt to filter out any spiky noise. I haven’t put my oscilloscope on the rails, or made any great effort to try and reduce any noise, but the SEM seems to be behaving itself so far.

I made some recordings of the SEM-in-a-box, hopefully it sounds a bit like it should. It’s the usual MC-4-driven frenzied sequencer nonsense.

Low-pass!

High pass!

Band pass!

Also I did a little recording with it along with the 606/Machinedrum for drums and System 100 for bass, and then thought it was boring so I started overdubbing a few more passes by hand. Like I say, this is just a temporary box for testing, it’s not going to have the Oberheim logo on it or anything when it’s done.

Tagged under ,

13 comments

  1. 3rd January 2022Al says:

    Hello! thats brilliant, and the samples sound excellent. Ive cloned about 4/10 of the SEM, dont think i have the time for more breadboarding, lol. Congrats!

  2. 21st June 2024JRC4558D says:

    Hi man, amazing clone! and sounds really, really nice, congratulations for the hard work!

    Over time I have seen several of your projects and have most of them saved as references :).

    I found this one just gathering info to start a SEM build exactly as the one you did, with all the components in the right place on the PCB. I f**king love it !!!

    Is there any chance of buying from you one PCB or the gerbers so I can send them to factory in case you have non for sale or not wanting to get involved in producing more?

    Thanks a lot for sharing, anyways!

    PS: love the dark electronics mood of your demo on youtube 🙂

  3. 6th July 2024ua726 says:

    Sorry for the late replay – thanks for the comment.

    I’m not going to be releasing any PCBs or Gerbers from this for now. There’s a chap on the Modwiggler forum who is making his own SEM four-voice and selling PCBs, you be better off checking his stuff out to be honest, I think he actually knows what he’s doing.

    Having built up my four SEMs into a frame with a basic panning mixer, and tried using them in a polyphonic way with a homemade poly MIDI to CV converter I have to admit that I’m not very excited by the sound of it. Getting the filter cutoff calibrated across them all is a bit of drag, and tuning the oscillators took forever.

    It also doesn’t help that the MIDI-to-CV converter I built outputs positive voltages only, which means that the oscillator pitch ends up being an octave or so higher than I’d like, despite the octave switches I built into the panels.

    Everyone seems to love the sound of the Four-Voice, so maybe I’m just doing it wrong, but in terms of my own clones I much preferred the sound of my wonky half-built Jupiter-4, which is at least a polyphonic synth from the ground up, rather than a bunch of monos strapped together.

  4. 30th July 2024JRC4558D says:

    Hi mate, thanks for the reply. Just reading it now.

    No worries, I’ll try my own way to do it. I found an online website that allows you to take measures from pictures (with some minimum error), so started to do that before loading the components in eagle.

    About modwiggler, you mean the guy called analogmonster? he’s doing a surface mounted version (I hate those little things) not interested in surface mounted components, those are just for comunications and high computers, there are not any good capacitors in that format so I’pass on that.

    Not liking the sound of the SEM 4F? it might be something about the tunning, the MIDI to CV or that is not the sound you like…

    Jupiter 4 is my fav of all the jupiters, I believe the filter is the key, bat then the VCO’s are also very unique as far as I remeber, I tried to buy one but it’s impossible unless you are rich.

    Anyways, again thanks a lot for the reply and I hope you find a way to make your 4 voices sound in a way you like and deserves the effort put in the project.

    I have a MU format modular all made and etched at home and the MIDI to CV is an old one but works really well, is not online anymore but I have the code for the micro and all the data needed to built one, just needs a good multimeter with at least 0000 resolution to be more accurate, if you have such multimeter it might work for you, let me know if you are interested by email and I’ll upload all of it to google drive.

    Cheers!

  5. 10th October 2024julien says:

    bravo for your hard work ! what’s your method to clone PCB ? I presume you start from the service manual and pictures and a CAD software

  6. 12th October 2024ua726 says:

    Yep, that’s it really.

    • Schematic drawn out in Kicad
    • Export netlist, then convert netlist/parts formats that OsmondPCB can deal with
    • Grab circuit layout from service manual (or photo of the board)
    • If there’s no parts designation on the board images (hello SEM) spend ages working out which resistor is which
    • Correct distortion and resize circuit layout to real-life size
    • Import circuit layout image as PDF into OsmondPCB as background
    • Import netlist/parts into OsmondPCB
    • Place parts, trace out circuit, adding twiddly bits as per original 70s design until I can’t take it any more
    • Correct any errors in the schematic based on the layout, re-export netlist/parts from Kicad/reimport into OsmondPCB
    • Export Gerbers, check the files in a Gerber file viewer

    It takes forever.

    My later clones have been closer to the originals as I’ve got more obsessive about it. The sawn-off Jupiter-4 motherboard I did had the same LFO bug as the original because of the layout.

  7. 10th December 2024LM394 says:

    This one has stumped me – cannot get the VCO to work – how did you deal the the +18.5 and +15 supplies needed? Also did you find any schematic errors – especially in the VCO core area – many thanks.

  8. 10th December 2024ua726 says:

    Huh, weird. As far as I know, the Oberheim published schematic is correct. The way I make my clones by tracing over the existing board layout tends to shake out any schematic errors, and I didn’t make a note of any for this one. I’ve just given my schematic a quick look over, and the VCOs look the same as the published one (save for getting the orientation of Q7 and Q16 wrong on mine).

    For the power supply I’ve been using an Oakley Sound Systems PSU board, like this one (but an earlier version), with a Yamaha PA-20 adapter. The PSU board can be adjusted via a trim potentiometer to run at different voltages, and mine has enough range to run at 18.5V, so that’s what I’m doing. There might be a way of altering the circuit to just run on 15V but I never looked into it.

    This power supply actually runs all four SEMs, plus a little panning/mixer output board, although I don’t think it could take a lot more.

  9. 10th December 2024LM394 says:

    Hi UA726 – appreciate the response – all other parts VCF, EG, LFO, etc… work fine. Using the same parts as you 2N3904/6 and J112. I built the regulators and that provides all the voltages. Done the breadboard over and over – no joy. Looking at the PCB layout – it appears to me that Q2 (FET) goes through a 2K7 resistor, when it should be connected to Q4 (PNP)? Q4 also seems to be connected to an 820K resistor which should be connected to IC3/Pin 2. Any thoughts or am i just reading the PCB traces wrong? I really wanted to get this one to work.

  10. 10th December 2024ua726 says:

    Yeah, here we go:

    Annotated layout for SEM

    The right-hand side of the FET Q2 here is attached to 18.5V, same as one side of R22 (the 2K7).

    Q2 is connected to Q4 through R20, which is the 1K5.

    At the point where the Q4 is connected to that 820K (R13) – that’s ground.

    Just had a look at one of my boards to make really really sure I was using a J112, and yep, they’re in there.

    Worth double checking you’ve used the right sort of transistor for Q4/5/6/7, PNP vs NPN – it’s the sort of thing I’d get wrong.

    Good luck with it. At least you can narrow it down to between Q2/Q4 and Q7 on the schematic, I feel like it’s going to be in that bit, at least. (Alright, it’s quite a big bit)

  11. 10th December 2024LM394 says:

    Yup – been a bit of a div. Can see those parts are tied to +18.5 and Q4/820K are tied to GND. Going to build another version with Japanese parts and trace the PCB as well – those have worked on other projects. Thanks again.

  12. 11th December 2024LM394 says:

    Amazingly – got this working. Turned out to be a bad batch of 2N3904/6 – gain was way to low <50 (shoud be around 200). So will be having words with the supplier. Glad I got this one going – was running out of ideas (and hair).

  13. 11th December 2024ua726 says:

    Hats off for sorting that one out, I’ll bear that in mind for future as well.

Write a comment: