[blog] MB-6582 - an 8xSID MB-SID synth

[Wilba]

Well it was never going to be perfect at the top of the knob, but it's fairly well spread around, no bright spots anyway... you can see the D-shaft cos it's black. The only solution to that is replacing the shafts with clear acrylic or something, and if I was going to do that I might as well use hollow rotary encoders and mount a diffused LED inside the shaft  ;D but I'm not that much a perfectionist!

I connect the LEDs with three in series and a current limiting resistor to the ground pin on each encoder, the resistors are mounted above the encoder pins, tucked under the little protrusion above the pins. I use the 11.3v from the input side of the 7809 (i.e. the 9v AC supply after the bridge rectifier/capacitor) and with a 3v drop per LED and 220 ohm resistors, this works out to 10.5mA per 3 LEDs, or 156mA.

The cooling fan is also supplied by the same pre-regulated supply, so I'm probably drawing 300-400mA before the regulator, but the draw on the 9v supply is so low that it doesn't really matter.... according to the 6582 datasheet, the SIDs only draw a maximum of 40mA each from the 9v supply.

[Jaicen]

So is it running yet?? I really want to hear a sample of it running 8x6582's!

Also, if it's 100% working, when are you planning to release the Gerbers etc...?

Personally i'd be willing to make a donation to obtain them as it would save me hours of wiring, which is my main pet-hate right now  ;D

[Wilba]

It's certainly 100% working, with the PIC18F4685 version of the MB-SID v1 firmware and also with the MB-SID v2 alpha firmware  ;D  but as I said before, I will release the next revision of the PCB layout and people can do what they want... it will actually work out nicer as the space saved by removing the IIC MIDI modules will be used for a 12v and 9v supply and two power rails above the SIDs so people can mix and match 6581 and 8580/6582... my board only has a 9v supply as I don't have 6581 SIDs. Also there's the CAN bus wiring which I've added to my board but should be added to the PCB layout. All these little extras will mean anyone who uses it won't need to muck around with a lot of fixes, just solder the bits on and go.

The CS is done too, and all works, but it's just not quite right... I think the LEDs are way too bright and perhaps too much blue, and I'm also going to change the knob illumination to use white LEDs (and very, very low brightness, just enough to make it visible but not so bright that the LEDs get lost). So the final pics of the finished box has to wait, sorry... I don't want to show off my mistakes! ;D

[Jaicen]

So the final pics of the finished box has to wait, sorry... I don't want to show off my mistakes! ;D

Hehe, I can understand that, particularly with your penchant for perfectionism ;)

However, I would say that a lot of people would benefit by hearing about your mistakes, so that we can avoid them in the future, so perhaps you could make up a bloopers reel of all the really juicy ones! (like the 8 PCB's I etched that were transferred backwards  :-[ ).

Could you give us an idea of how much the three PCB's will cost to get manufactured?? I'm in the UK, but i'm sure the costs will be around the same (accounting for exchange rates). I'd be willing to get a group buy together if there were more than one person interested in this project (and how can you not be, it's everything you could need in one tasty box!).

[Wilba]

I posted about costs in an earlier post: http://www.midibox.org/forum/index.php?topic=8389.msg62100#msg62100

It is obviously going to be a lot cheaper if you wait for a batch made by SmashTV... and if enough people really want a control surface identical to mine (and I don't think many would because it's a matter of personal taste, I went for minimalist white on black and really compact, others prefer more space and multi-colour artwork, etc.) then perhaps you can batch order the panels from Front Panel Express too and save heaps.

I just don't want to release the layouts before I've done a revision and added the missing bits, taken out the redundant bits, and put in the bits TK suggested, i.e. the things I changed or took out because I didn't need them and was building it for me not for everyone  ;D  eg. the circuit for controlling LCD backlight brightness is not on my PCB as I use a PLED display. It's also easier for me to not spend a long time describe how I've "fixed" it to the new MB-SID V2 hardware changes, better that I spend that time fixing it on the layout and have less documentation and support issues (yes, I will be "supporting" this PCB and answering questions people have).

I can't list many mistakes in this project, as there weren't many! Here's some advice:

The #1 thing that helped this project was having so little free time to actually do stuff meant I spent a lot of time thinking about it, planning it in my head and allowing enough time to change my mind on how I was going to do things. It's not some skill I can easily transfer, the advice is not to rush things and then you won't end up ordering a PCB that's got mistakes or having the whole thing not fit together.

The #2 thing was getting a cheap pair of calipers to measure components. You can trust datasheets most of the time, but sometimes you don't have datasheets... like how big to make a hole in a panel for the C64 power switch, and where to put the pads for the C64 power switch.

The #3 thing is getting the components, especially the case, in your hands before you order PCBs to fit it, and making sure things will fit, with printouts of the PCBs. You can catch many mistakes that way, just by putting the printouts in the case to see how it will fit, or overlaying the front panel printout with the PCB printout (I caught misalinged switches that way!)

OK, so there were a few mistakes:

The height of the display was 10mm in the datasheet but was 11mm in reality. Not that big a difference except when you're fitting it in the 10mm gap between panel and PCB. That was easily fixed with a routed hole in the PCB, and I planned for that possibility too, if I needed a bit more space between the panel and the display's PCB (that's where the extra PCBs go with the select and SID buttons).

Where the CS PCB and base PCB come close together at the bottom edge of the case, there's not enough space for SIL connectors (like on the bottom side of the PIC) and even the IDC connector is a tight fit, I had to remove the extra little bit of plastic (the bit the cable normally gets tucked under as stress relief). Since I only used those SIL headers for testing and uploading (it's the DIN/OUT port and Tx/Rx port) it wasn't a real mistake and I knew it was going to be a tight fit before I ordered the PCB. But in the next revision, I could probably move the 8x BankStick section up a bit and move all the Core sections up too and give more vertical room.

I assumed I could neatly shorten one of those ALPS tactile switches and put them where they needed to be shorter in those extra PCBs above and below the display, but that didn't work out too well. Luckily they make them in just the right size to fit there too.

I don't know if this counts as a "mistake", but while I really like this case, panel and the PCBs, it was a tight squeeze to make everything fit, especially the lack of room near the bottom of the case where the PCBs get close. It would have been easier to build it in a slightly bigger case, less challenging I suppose, but more accomodating if I needed more space between CS PCB and front panel, for example.

And the final mistake (which I'm fixing now) is the LEDs I have are too bright, even when multiplexed in the switch/LED matrix (so all LEDs are only on 1/8th the time) with a 220R resistor on the DOUT outputs it's still way too bright, and I need to increase them to 2K or maybe even 3K-4K... that will make those transistors on the current sinking DOUT a bit redundant if I'm only sinking a max of 12mA through the current sinking shift register! The illuminated knobs were a bit of fun that I planned for (note the oversize holes in the panel) and while it works fine, I think they also are too bright, they'll be much better if they're white (so not confusing with the panel LEDs) and just a faint hint of light coming off them, just enough to give some contrast with the black of the knobs so they don't look like they're all black.

I'll post the final pics when I've toned down the LEDs and swapped in some white LEDs under the knobs... still waiting on the white SMD LEDs  >:(

[Wilba]

Final construction shots (for those into gear porn)  :o

Link to my Flickr page

I finally wired up the last little bits, the four dual-gang pots at the back (for audio feedback), the passive mixer to stereo phono jack, increased the DOUT resistors to 3K, added a pot to turn down (off!) the illuminated knobs, screwed the PCB to the case and it is now all finished!

I've eventually decided to turn off the illuminated knobs. One knob by itself does look cool, 15 of them mixed in with the LEDs of the control surface is just too busy, even when I turn down the brightness. Oh well. Maybe one day I'll get those SMD white LEDs and give it a go, for now I'm just going to leave it off.

OK, time for some sleep, then I might take some shots for the MIDIbox of the Week  ;D

[dj3nk]

week ??????????????????

It could get "ever" I say ! Incredible work what you have done with that little space! Do you already know if and when you want to publish the pcb files? What was your hole cost for it? (with casing etc)

greetz

[stryd_one]

Yeh this is up there with the best for sure. Freakin awesome Wilba!

When you gonna bring her over? I wanna touch it :-* ;D

Edit: I wonder if there might be room in v2 for controlling digipots so that feedback could be saved in a patch?

[TK.]

I wonder if there might be room in v2 for controlling digipots so that feedback could be saved in a patch?

this is already possible - there are 8 16bit values implemented for customized external (or internal) parameters.

They can be controlled via modulation matrix, via knobs, via MIDI, via wavetable...

Best Regards, Thorsten.

[Wilba]

Now that it's finished, I can turn my attention to the PCB layout, make some small changes (size and component positions will stay the same), and depending on the level of interest, SmashTV might get a batch made and sell them, or if that doesn't happen, I'll just release the Gerber/NC Drill files and anyone else can get a one-off made.

I find it hard to calculate the total cost, as I'm always buying more bits than I need...

As a rough guide:

Case: AU$45

20x4 PLED display: AU$45

LEDs, switches, encoders: AU$70

Knobs: AU$20

PCBs: AU$170

Panels: AU$200

Other components (chips, sockets, screws, spacers, etc.): AU$100 (I guess)

8x 6582A SIDs: Priceless  ;)

So that's AU$650 without SIDs, but I'm probably forgetting a lot of little bits that were bought, a few salvaged C64s, a blue LED 40mm fan (from a Northbridge cooler)... it's probably closer to AU$700 with another AU$100 for the SIDs.

AU$800 (US$662, 486 EUR) is a lot, but it's cheaper than an unavailable SidStation, with eight times the SIDs and about a million times the synth capability. I don't know how much an x0xb0x costsA lot of the cost is in the prototype PCBs and panels - buying a batch run PCB and using the aluminium front panels that came with the case would be a lot cheaper.

Plus when it comes to MIDIbox, I'm spending the money to have fun while building it, and to build something exactly the way I want it, so in that respect, AU$800 over the past year isn't a lot at all.

[stryd_one]

there are 8 16bit values implemented for customized external (or internal) parameters.

They can be controlled via modulation matrix, via knobs, via MIDI, via wavetable...

:o OMG that's awesome! You're really working some magic on these new PICs. Sometimes I think if you got another 64k of code space you'd become immortal ;)

[ptitjes]

Hi Wilba,

I want to do my own board for my panel components. I just run along formulas to calculate copper sections depending on resistances but i'm not sure if i applied them correctly. Is 9 mil sufficient for signal tracks or could i just do the 7 mil minimum of Gold Phoenix ?

Could you tell me what trace and clearance size you used for both *Vss* and *signal* tracks ?

Thanks. Didier.

[Wilba]

I use a ground plane on the bottom layer, so I don't have many tracks of ground, the few I do have are for the SIDs, which I'll explain later.

I've used 10mil for all digital signal tracks and short little "branches" off the digital +5v "rails" (eg. to a pull-up resistor, or to hold a chip pin high). The +5v "rails" are clearly visible in the PCB layouts, as they are 30mil, from the PSU section in the top-right corner, it travels down the right side, then splits into two rails, one going under all the PICs, the other going along the shift registers at the bottom edge.

All the "audio" tracks for the SID, including tracks to connect the SID's ground to the "audio" ground plane, and the +5v supply to a separate +5v supply rail, are 30mil.

I've used 10mil clearance throughout, although I manually give the 30mil tracks a bit more clearance than that, not that it matters much. All vias are 50mil with 28mil hole diameter.

Essentially that's all really fat by modern standards, as you would notice just looking at a modern PCB like an LCD module or mobile phone... for digital stuff like this, you could go as small as 7mil, especially for just a control surface board where your currents to LEDs and switches are so small anyway (a few mA at most), but I chose 10mil for convenience (fatter than any PCB manufacturer's minimums) and on the chance I ever needed to fix a track, I'd have a better chance of fixing it  ;D  Same deal with vias, they're fat but I wouldn't have a problem getting it made. If you're doing mostly PDIP and through-hole, 10mil on two layers is more than adequate.

One piece of advice, don't go too small on the pad sizes, I sucked one up while desoldering, so if possible, make it a little bigger. 60mil round pad with a 35mil hole is fine to solder once, but desoldering can be a bit of a problem if you have to... so go up to 80mil round pad, especially if you're planning to solder wires to it that you think you might desolder once or twice.

[seppoman]

Hi Didier,

"Warning - while you were typing a new reply has been posted." - anyway :)

here´s a link with common design tips - it´s from Olimex (Bulgaria) where I ordered my last PCBs, but the tips are generally useful:

http://www.olimex.com/pcb/dfm.html

About track widths: Don´t push the widths and clearances to the limits of your manufacturer. The narrower both become, the higher the probability for manufacturing errors. For a standard PCB with only normal through hole components, there´s no need to go below 10 mil. I use 16 mil width and isolate for normal signal traces. as long as you don´t need to have traces between legs (which shouldn´t be necessary for a panel board), you could even use 24 mil. Generally, the wider the tracks, the less hassle and thoughts to draw into account. For supply traces, I try to use at least 36 mil. If that´s too much near an IC or connector, I use a narrower trace just for the last part forking from the "main trace". For parallel analog or clock signal traces, it doesn´t hurt to try to have them separated with a small strip of ground plane in between.

Seppoman

[ptitjes]

Thanks guys for your quick answers! Wasn't there this afternoon to read so some more questions...

Sorry if that are stupid question and if that is off-topic... I'm a very newbie!

Wilba:

Why would i have pads ?

Also it looks like your Vss on your surface board is thinner than on your main board ? Is that 20 mil ?

Seppoman:

Very cool link! I'll read that up when i'll had more rest... :)

Both:

I finnaly succeed to have only one layer by better placing my sil connectors. Do you have recommendations on track sizes for homemade boards or should i stick to the 10/30 ?

Thank you very much for your help!

Didier

[seppoman]

Do you have recommendations on track sizes for homemade boards or should i stick to the 10/30 ?

Try to use at least 16 mil traces and 12 mil clearance - I´m not saying 10 mil traces are bad, but in my opinion you should have a reason for choosing narrow traces. For power traces choose 24 or 32 mil. (of course 30 is just as good, but 12 - 16 - 24 - 32 are standard widths in Eagle).

Seppoman

[Wilba]

Sorry if that are stupid question and if that is off-topic... I'm a very newbie!

There are no stupid questions, just stupid people  ;D

(disclaimer for the sarcasticly impaired: I'm joking, I'm not calling you stupid!)

Why would i have pads ?

... although not a stupid question, it's a bit confusing! Pads are where you solder component pins to the PCB. For through-hole components, you need a hole through the pad, and the pad is then a ring/donut shape. Keep in mind also for some parts like rotary encoders, they have "snap-in" flanges on the sides, so you can put large holes to fit them in, and then bend them on the other side and/or solder them for stability.

Also it looks like your Vss on your surface board is thinner than on your main board ? Is that 20 mil ?

Good catch... I like someone who points out my mistakes!

I started writing my reply about the main board then later realised you're asking about a control surface board. You're right, on the control surface board I use 20mil traces for the Vss, which is only connected to the encoders, as everything else is in a matrix and is sinked by DOUT pins. I probably could have used 10mil there... it is all very low current so 10mils would have been fine.

Basically you now know things will work at 10mil without a problem, but unless you've got some really tight layout that requires 10mil, try using thicker traces.

[ptitjes]

There are no stupid questions, just stupid people  ;D

(disclaimer for the sarcasticly impaired: I'm joking, I'm not calling you stupid!)

There is no offense. I was always told that people not daring asking questions might be the stupid ones... :P

... although not a stupid question, it's a bit confusing! Pads are where you solder component pins to the PCB. For through-hole components, you need a hole through the pad, and the pad is then a ring/donut shape. Keep in mind also for some parts like rotary encoders, they have "snap-in" flanges on the sides, so you can put large holes to fit them in, and then bend them on the other side and/or solder them for stability.

Ok. I thought pads were called pads only for SMT and vias for through-hole components...

As for the "snap-in" flanges of the encoders, i've done gEDA (pcb) elements and i've put big "pads" on them for the flanges. However, maybe i could put more copper and cleareace around them if i want to solder the flanges...

Good catch... I like someone who points out my mistakes!

I started writing my reply about the main board then later realised you're asking about a control surface board. You're right, on the control surface board I use 20mil traces for the Vss, which is only connected to the encoders, as everything else is in a matrix and is sinked by DOUT pins. I probably could have used 10mil there... it is all very low current so 10mils would have been fine.

Basically you now know things will work at 10mil without a problem, but unless you've got some really tight layout that requires 10mil, try using thicker traces.

This last sentence does remove my last doubts!

sepoman: I use gaf (gEDA and friends) and default values seems to be a little thinner.

Thanks for all guys. Going back to layouting... :)

Didier.

[rutgerv]

Hi Wilba,

just curious...what kind of powersupply do you use to run this 8xSID module on? Do you know anything about the powerconsumption of the board?

The ucapps.de website used to say '1A PSU' for the MB_SID V1 with 4 SID's and control surface step C. Is this still valid for the new generation of 8xSID MidiBoxes?

Greetings,

Rutger

[Wilba]

I use the C64 PSU (big black external transformer) which generates 9v AC at 1 amp and 5v regulated DC at 1.5 amp. This plugs into an original C64 power socket ;) There was no chance of fitting the required transformers into this case, and having it external is a neater solution even though it is a huge transformer!

The drain on the 9v supply is not too bad... it is 25-40mA per SID, plus whatever the audio buffer circuit consumes (not that much I think), and on the 5v supply, it's 70-100mA per SID, although I can't see how it gets up to 100mA, maybe if it was switched to output on the data pins, or maybe it varies based on the oscillator frequencies or filter use (there's an internal resistor ladder made of MOSFETs for the filter control).

So even in a worst case scenario, 320mA on the 9v supply, 800mA on the 5v supply... but you have to be careful because the "optimized C64 PSU" circuit adds the 9v to the 5v to generate 14v, and then regulates it back down to 12v or 9v... so some of that 320mA could be coming out of the 5v supply too (not in my case because I just regulate the 9v AC from the C64 PSU directly).

[rutgerv]

Hi Wilba,

thanx for the info! I still find it quite hard to make a good estimate of the required amount of Amps of the powersupply in my case.

I've designed a PCB with:

8x SID

4x CORE (1 of them with LCD-connection)

2x DIN (for control surface)

8x 'bankstick' EEPROM

4x AOUT_LC (each has 4 x 74HC595)

8x Moog Ladder filter with added CV control for resonance

1x Mixer/Output stage/Headphone buffer

I left out the power regulation circuitry for each module because I plan to create one powersupply that delivers all required voltages. These are the powerlines I used in the design:

5V - analog (SID's)

5V - digital (all digital stuff)

9V    (SID's ....a jumper selects the proper voltage for the type of SID)

12V  (SID's and all op-amps ....a jumper selects the proper voltage for the type of SID)

-12V (op-amps)

In my opinion there are two options for a global power supply:

- One huge dual 12V transformer and regulators to deliver the 9V and 5V

- One dual 12V transformer (the 9V will be downregulated) and one 5V regulator

In both cases I'm wondering what the amount of Amps for the supply should be...

Greetings,

Rutger

[Wilba]

I can't really help you with that... I don't know where to begin with calculating current requirements of the filters or op-amp stuff... I've been getting along just looking at max current values in datasheets.

I do know that other people have worked on a circuit for supplying mixed voltages like you want, and dealing with the power drop from 12v to 9v and 5v... check on the forum, there's some good info there about how to put a high power resistor to drop the voltage without overheating the regulators. That's essentially your biggest challenge, how to use a 12v 2A transformer (which will be 12v at full load, and much higher when less loaded) and converting that into the lower voltages.

[malekko]

Wilba,

I'm speechless, you did such an awesome job on this project!

I look forward to getting PCBs when they're ready...you mentioned you have some 6582s FS...could i put myself down for 9 (it's good to have one spare)

and i missed where you got those knobs from?

now back to drooling at the pics!

[Wilba]

So many people ask about those knobs in the other forums too.

They're from Alltronic/ALBS. None left exactly like this, but the red ones like on TK's gear are still available. I'm sure I've posted about these knobs a dozen times at least...

These knobs are available from ALBS: http://www.albs.de/

part #863062, "Drehknopf DK16-190V3 A.6/4,5 AT=14,5 schwarz soft-transparent rot TR-3925"

http://www.albs.de/ecom/images/863062.pdf

All my spare 6582 SIDs have gone on sale and been sold. I'll add you to the list of people wanting them if I get more, but there's no certainty I can!

[malekko]

thanks for the info.

I really like the sound of the 8580, but i only have 6581s...i cant ever seem to get the 8580s though. And to be honest i've never even heard of the 6582 until today. I've been doing research on them every since i saw your box.

Thanks again for posting that great looking box...the LAST  thing i need is another project...but as soon as the PCBs are available i'm putting everything else on hold!

now to find 6582s or 8580s... ;)