Self syncing VCOs

[skee]

Happy new year all. I completed my build of the Soundtronics midi Ultimate synth last year and with help on this forum greatly improved the stability of the three VCOs, which were suffering from temperature drift, brought on by the positioning of the expo transistors. Over the holidays, I have had the opportunity to play around, and noticed that when tuning to unison, instead of the sound becoming richer and interesting, it went flat.

Looking on the scope showed that when the traces approached within a few hertz, the signals would lock together, in frequency and phase, and stay locked for several minutes. To overcome this, I split the pcb power lines away from one of the VCOs and inserted new, separately regulated +/- 12 volt supplies. This doesn’t seem to have had much effect and the locking issue persists. The synth is otherwise working well and looks great. The VCOs are basically a somewhat cut down version of Ray Wilson’s mfos design. The actual sync inputs are not connected and are open.

Anyone got any suggestions on how to rectify this.?

[devinw1]

It could very well be a layout issue. Think about actual physical location of each VCO, and also look at ground paths, and any parallel traces that have each VCO's signal. Remember that parallel traces act like a capacitor and can cross couple signals. You are wise to look at cleaning up the power to each VCO too.

[guest]

oscillators really want to sync, and to get them to not sync is difficult. inside of some synths they actually put a metal cage around each oscillator, to keep it from radiating and syncing with its neighbor. so it could be through the I/O, through the power/ground, or just through the electric fields in the air. it might be good to start narrowing things down, and the powersupply test is a good one. did you put a capacitor on the board where the powersupply connected to it? long leads from a supply to the board will pick up RF. can you get 2 oscillators to sync, while keeping a third well enough away that it doesnt sync? can you get any combination of two to do this, or just adjacent ones? is modulation turned down to 0?

[KSS]

What the experts above said.

Can you share a link and-or PCB pics?

[skee]

Thanks for the replies. The oscillators lie beside one another on a single pcb, along with the other modules that make up the synth. I’ve taken a photo and I’ll upload tomorrow. I’m using 1 and 3 to test. Two is in between and the power lines have been cut in preparation for connection to a new regulated supply, so 1&3 are separate, apart from 0V, which cannot be separated. It looks like 1 is syncing to three, but it’s hard to be sure. The sync lines are connected to ground by 100k.

There are two 100uF caps close to where the new power lines were connected

Tonight’s efforts centred around pulling the chips from VCO2, the mixer and the noise generator, but no noticeable improvement. I’m curious that a kit that was published in a well known magazine and supplied by an established supplier appears to have flaws that appear to be embedded in the design, and weren’t caught before offering it to the public.

I’d like to hear from any other constructors of this kit, and how they have fared to see if it the design or my workmanship that is at fault.

[KSS]

I found these photos on their site. <--Hate having to do the groundwork to answer a question.
The PSU appears to be a very simple 78xx type with minimal filtering. Once you share your photo of the built VCOs we'll be able to see more. But it's by no means unusual for designs to have issues like these from even well established mfrs like Roland and Korg. And even moreso now when EDA has progressed to the point that -apparently- many feel it does all the necessary work for them.

You say you have no access to the 0V and yet that is the single most common thing causing this kind of osc. locking.

To give useful guidance beyond general principles, it will help to have detailed photos of both sides of the PCB and VCO schematics. I didn't take the time to see if these were somewhere in their support forum. That's your job IMO. But we'd be helped to aid you by having them and good photos. The PSU PCB pretty much shows its circuit.

[KSS]

There are two 100uF caps close to where the new power lines were connected

Look for other de-coupling caps of much lower value. Near the ICs and also at the PCB entrance near these. If these aren't present, they could be part or all of the problem. I'd add a couple to the PSU also. But we can cover that once we have better photos and hopefully VCO schematics to work with.

EDIT: Standard decoupling caps *are* on the PCB and in the schematics. Mostly. But we'll still be interested in the details of their traces. And possibly their values too.

[lemberski]

The schematics are here: https://www.soundtronics.co.uk/index.ph ... ent_id=444

[KSS]

I found some photos of the VCOs. The method shown for the expo pair and tempco is not good. This may or may not be affecting your sync, but it needs to be addressed either way. Instead of putting the two transistors as the silk shows with the tempco between the rear of one and face of the other..

Put the tempco in low and unsoldered. since yours is likely already soldered, heat one lead at a time and while liquid, push it down towards the pcb. then the other lead.

With the tempco now low, make the left expo transistor do an extreme 90 degree back bend and the right -rear- transistor do a similat extreme front bend so that their flat faces meet in a tight embrace. Put thermal compound between them. This alone will make your VCOs more stable with temp changes and more evenly balanced in the expo output when frequencies change. Bring the tempco up against the lower rear of the leftmost -now lowest- transistor.* This can be essentially ignored if you put the entire three components under some king of 'box' enclosure. They all need to 'live' in the same micro-environment.

*see the later post below about whwre to put the tempco body and how to build an insulating box for all three parts. /edit

Now, none of this is 'directly' linked to a sync issue. But please put it on a high priority list of things to do ASAP.

Thank you, lemberski, for the schematics. Between those and the layout above, it's safe to say the person who laid this out was not an experienced synth electronics designer. <--Doesn't mean it can't or won't be a great synth. Only that the marks of inexperience are all over it.

And strengthens the original reply that its layout may be the most likely cause for your issue.

The Photos I found are not clear enough to clearly see the traces around the sync-reset transistor -on the PCB.

Are you able to read and understand schematics? Sheet 2,3 and 4 from the zip are the VCOs.

Can you give is a clear photo of the region around the 8 pin DIP socket and its nearby transistor? Both sides of the PCB.

[KSS]

This might sound crazy, but it's possible that reversing the pins of the PS timing caps can make a difference. Older polystyrene caps were labeled to help get this right. One leg is tied to the outermost foil layer, which will make it more susceptible to outside interference. But these PS caps are also *highly* sensitive to over-heating when soldering so *don't* make the change willy nilly. For now, just keep in it the back of your mind as we explore what might be happening here to cause your unwanted OSC locking-syncing.

Something you *can* easily do is to stand up the PS timing caps of VCO 1 and 3 -assuming VCO2 is still taken out of the situation- to get thenm as far away from any other component as you can without un or re-soldering.

AIRC MFOS OSCs were rather prone to this locking -and general instability- from the soundlab days too. Which means you might find some answers in some *old* support threads for those much earlier MFOs products.

[KSS]

I've converted sheet 2 to a JPG to more easily share it here.

the MPF-102 is the reset device. I don't remember if those have equal-interchangable source and drain?
Will need to check the datasheet once you tell us which transistor mfr made the ones in your VCOs? So we get the right datasheet.

I've colored the components used or corollary to the sync operation. Pink for passives and blue for semiconductors. It's not perfect, but this kind of highlighting can make a schematic MUCH easier to follow. It should also help you when looking at the PCB for possible trace interferences, soldering, and so on. Does the theory of operation of the sync make sense to you? I didn't color the timing capacitor. But we understand it's going to be crucial to the sync-reset for the VCO.

Suggest you fix the tempcos and expo transistor pairs. Then stand up the PS timing caps in the active VCOs without de-soldering them. Let us know what you have at that point.
-----------
All of this advice based on *your* VCOs and PCB matching the ones in the photos I found.

[KSS]

One more thing.
Can you share how long the cable between your PSU box and main unit is?
And what gauge wire you used between them.
Which connector-s?

EDIT: Deja-Vu on this PCB layout and expo. I participated in another MW thread not too long ago where I remember suggesting building a thermal insulating box around the expo pair and tempco using the trimpot body as one wall and as support for the two side walls and top. Putting the tempco body between the pair and the trim to use the trim as a 'windscreen' even if the whole 'box' wasn't made.

[jorg]

One more thing.
Can you share how long the cable between your PSU box and main unit is?
And what gauge wire you used between them.
Which connector-s?

EDIT: Deja-Vu on this PCB layout and expo. I participated in another MW thread not too long ago where I remember suggesting building a thermal insulating box around the expo pair and tempco using the trimpot body as one wall and as support for the two side walls and top. Putting the tempco body between the pair and the trim to use the trim as a 'windscreen' even if the whole 'box' wasn't made.

It would be even better to replace the separate transistors with a pair such as BCM847. Their thermal coupling will be much more efficient.

https://assets.nexperia.com/documents/d ... _BS_DS.pdf
http://www.proto-advantage.com/store/da ... PC0130.pdf

[devinw1]

Yeah but it sounds like this is a pre-made kit so adding SMT matched pairs to it may be behond the scope of what the OP can do or is willing to do.


Anyway, If it does turn out to be related to the PCB layout, which it very well may be, I will warn you OP that there will not be much you can do about it without getting pretty crafty. And by crafty I mean spinning your own boards for the VCOs that connect back into the main board at the correct patch points. If you can't solve the issue with the existing PCB design with component level changes (decoupling caps etc), then you'll may up pulling your hair out trying to cut and replace traces and do other board surgery trying to solve the problem and likely won't succeed. It's usually more efficient to design in from the beginning, at the board level, to make sure the oscs don't talk to eachother, but of course you didn't design this so that isn't your fault nor something you can easily edit and re-spin the boards. Not trying to be negative, just preparing you for the possibility!

[skee]

I’ve attached a photo of the three VCOs on the board, but I don’t think it will add much to the discussion, over what has already been unearthed. I can access the ground, but it is “inaccessible” by virtue of being a ground plane, rather than a single line or lines that can be separated from the rest of the circuit. The leads connecting the psu box to the pcb are about 8” long and are seven stranded. The VCOs under test have about four inches between them, with VCO 2 (temporarily disabled) between them.

I don’t like the power supply provided, which is a single 12 volt AC source, using half wave rectification to split it into two 17 volt +/- rails and having large reservoir caps to compensate for the lower efficiency. I suspect this arrangement of being on the noisy side, so I plan to replace this with a split 12-0-12 transformer and full wave rectified. Maybe this will help.

Sadly, I have almost resigned myself to the separate VCO board solution. My real peeve is that I had researched this quite thoroughly, and opted for this solution, which was not cheap, in order to avoid this kind of issue.

[guest]

you can try a tinfoil wrap around one of the VCOs, and see if that changes things. youll need to line the udnerside of the tinfoil with something non-conductive, and then clip the tinfoil to ground.

as devinw1 mentioned, you can cut the pcb traces and re-run, them or put some more decouplings caps in. another thing you can try is grounding out some of the inputs or removing components and seeing if that helps. for example, remove R114 in the schematic above. ive seen signals capacitively couple through switches before. then put a big capacitor at the inverting input of U102.2. basically, its a lot of going around and decoupling anything that goes to another part of the synth, or to the powersupply rails. if that doesnt solve it, and shielding doesnt solve it, then you have to cut the ground plane.

[skee]

I’ll certainly try removing R114. I’ve also considered using a cheap voltage reference, TL431 in place of the R113/R123 divider. If this doesn’t work, and the ground plane can’t be separated, then it looks like I’ll be learning Kicad.
@KSS thanks for the research and suggestions, it will take a while to work through them I am probably responsible for your deja vue, and have successfully implemented the mods to the transistor pairs. To simplify the ‘bend and bow’ procedure, I filed one of the rounded faces flat (diamond file required) and mated them that way. The thermistor then fitted snugly in the resulting groove in the side.

[rich_de]

guest beat me to it with the resistor. And skee too, with the reference. I saw the latter after I typed and just before I posted, Haha. Anyway. I'll just second them. R114, R214 and R314 connect the wave peak reference voltage to outside. I've had the same problem in an own design in that location. I'm not entirely certain what caused the interference, but when I relocated the sync-forcing transistor with its lead to the reference far away from the other VCO, the self-syncing went away.

One thing you can also try is to supply the sync voltage from a separate reference (as suggested by skee), because relying on the supply 12V isn't a particularly solid solution. E.g. take a LM4040 5V with a buffer (or just a separate 7805, or mentioned TL431), take out R113/R213/R313 and feed the 5V through 3K, so you get the 2V from the divider like in the original.

For the transistor pair, if you want absolute stability, it looks like a DIP 8 LM394/AS394 pair can barely be wedged into the existing holes (notch down), with the two NC legs up or outside of a 6-pin socket, and the case scraping along the trimmer.

[KSS]

Please try the things I suggested before throwing in the towel.

You are also o nthe right track about the power supply improvement. "7 strands" leaves out the important part which is the wire gauge. What is that? And what are the connectors?

I put a lot of effort into providing a useful reply and it kinda sucks to see it disregarded. This is not my first rodeo.

[KSS]

I asked for PCB photos of both sides. Was that too much work? Detailed close-ups to show more than the ones I LOOKED FOR AND LINKED ON MY OWN TO HELP YOU.

Yeah, I'm disappointed.

Go ahead and go after all the other things though. Knock yourself out.

[devinw1]

You can lead a horse to water, as they say!

[skee]

If I wanted not to be listened to, given a long list of irrelevant jobs to do and then be shouted at that I am lazy and ungrateful, I don’t need to come to Muffs. My wife has that pretty much covered, thanks.

[devinw1]

Hahah, goddamn. That's pretty funny