Was really hyped when the Synth Pro finally hit the stores, not only because I like MFB as a company but also since we’re talking about a polyphonic synthesizer with great specs and sound. There are 3 VCOs that can be switched to DCO mode and have some other tricks up the sleeve, two different filters and very decent modulation options.
Some bits here are guesswork (flagged as such), especially as regards the VCO/DCO switching, but things are gradually evovling with each multimeter beep, so check back in again later. Also, if you have further pointers, do shoot me an email via the contact form!
Unmodded sounds are here: https://soundcloud.com/uibkmedan/mfb-synth-pro-sounds
Extra Programs for the Spin FX processor
DSMURRAY, an ace person dedicating time and energy to Synth Pro matters, has an extra eeprom for further FX programs on that machine: http://www.bitchintunes.com/spfx1
The same DSMURRAY also mentioned this PSU as a potential replacement for the existing one in a social media group: https://cpc.farnell.com/ideal-power/25hk-ab-060a300-d56/desktop-psu-6vdc-3a-c14-vi/dp/PW04186?CMP=TREML007-005&fbclid=IwAR1Q78ewcJCQolCbHysH5iRUJ2ga20jN9qa5OYy5VTefTL8TltqvxbAx1BQ.
First, take off all potentiometer caps. The shafts are T18 and these caps sit tight, so if need be, gently pry with some wedge/teaspoon (but mind dents in the chassis!). Take out the four hex-screws marked with green boxes on the picture. You can now take the top lid off. This is enough for fixing (if needed, see also below) the VCA release flutter thing some early units had. It would also be enough to require the headphone output into a FX processor input (not that this would be interesting…).
If you need to take out the entire innards for working on the voice boards, then first take the nut off of the main output jack socket. Then only unscrew the screws marked in boxes below. The other screws hold the voice boards fastened to the main PCB. Note, there are metal washers under the PCB and the chassis spacers holding the PCB in place (these washers are attached in a way that it’s hard to put them properly in place again). Also note that you cannot lift the PCB out horizontally. Main PCB and voice boards form a kind of sandwich, and the main power on/off switch protrudes a bit into the gap between both. Hence, lift diagonally (bottom up first, where the display is), then slowly pull towards you. Also mind the cables connecting PCB and on/off switch.
There’s one main board with CPU, display, FX processor, controls and I/O and on its bottom you find two voice boards attached, which comprise 4 voices each. [The first unit I had still sported some coolaudio V13700 chips, but on later units Lm13700s are used instead.]
Main PCB and Output Section
On the main PCB you have a couple of voltage regulators, your set of controls, two CPUs (one is under the display), and the FX processor (Spin) and output mixer. The 8 voices are summed at the voice boards – since the filters can operate in a stereo-spread configuration, you have two summing nodes (one for left, one for right), which are fed to the dry signal mix VCA (a stereo VCA controlled by the main volume pot) and to the FX processor (which has a stereo input). The FX output (again, stereo) is fed to another stereo VCA (controlled by the dry/wet pot). These VCAs are built around LM13700s. A quad op-amp (TL084C) is for summing dry+FX signals and forwarding them to the main output and the headphones out. There’s another dual op-amp here, one half of which carries a mono sum and the other half some broken sounding mono sum – not sure why (differential amp for limiting noise?//used for tuning routine?).
Release flutter fix
Some early units had slightly fluttery level fluctuations on sounds with a long release envelope (check gearspace synth Pro thread for more info/sound clips). MFB forwarded information how this can be fixed by changing four resistors on the final mixer (specifically, the dry mix VCAs). I did this on my unit, and it’s relatively easy to do. MFB offer to do this and DSMURRAY (yes, a very active Synth Pro person!) does this in the UK.
You can easily check if your unit is the updated/improved version. If resistors 1+2 (see pic) resistors on your unit are labelled “4701” and resistors 3+4 read “1003”, you might want (but don’t need) to change them.
Underneath the main PCB are two boards carrying each four entire voices. On the picture above I labelled two of these. The leftmost is labelled according to a picture Manfred Fricke posted on the sequencer.de forum, and on the one right next to it I labelled IC types. Power rails for ICs are +5v and -5v (per voltage regulators on the bottom left hand of each voice next to a LM13700 (VCF1). Another +3.2V regulator is next to each CPU. The CPU itself handles switching, modulation duties, etc. (and also provides the noise generator as far as I can see). What’s analogue are the VCO/DCOs, filters, Mixer, VCAs. So, in brief, you have a fully digitally controlled analogue synth voice.
Tracing this is a bit on/off work and my knowledge of temperature-dependent tuning is less then developed, so not all’s spot on here. Couple of folks emailed on this and I’m doing some reading to catch up on thermal drift matters.
Each VCO/DCO is built around a saw core (transistors and op-amp [TL064]) which is then shaped further to triangle (also at TL064) and pulse (at LMV344 op-amp). With two op-amps of each TL064 used for the saw core (A and D) and one for triangle shaping (C), we have one op-amp (B) left per IC. On VCO1, the remaining amp serves as a differential amplifier used in the filter mixing stage, and on the two others, the “leftover” amps are used for “ringmod” generation on VCOs 1 and 2. I put “ringmod” in quotation marks because – I need to trace this more – this seems more to be a case of differential amp + single diode setup than a conventional quadrant multiplier. Sub oscillator I still ned to trace/understand more too, and noise, as it seems, is generated digitally.
As for pitch stability on the Synth Pro, I think it’s useful to understand the exponential converters better, especially as regards thermal drift. A couple of users keep reporting on needing to use the tuning routine frequently, so here’s hope that this can be improved. Undortunately, I’m just beginning to read up on this and might not be the expert here anyways, to say the least. Still, I’m trying to describe this as best as I can and keep updating corresponding info.
The exponential converters for VCOs 1-3 are built around a single LM3046 NPN transistor array (with an additional single NPN labelled WCs 8N). The LM3046 sports five NPNs, with two of these already set up as a differential pair (Q1 and Q2). You can set up another pair (Q3 and Q4) in the same fashion, and there are indeed some classic synthesizers using such setups (https://modwiggler.com/forum/viewtopic.php?t=131832). In order to have six pairs, an aditional single NPN is used, yet, and here comes the catch – close matching of these NPN pairs is crucial for pitch stability, especially as rgeards thermal drift. There are various methods for compensating such drift, see for instance here http://electronotes.netfirms.com/s019 or here https://www.schmitzbits.de/expo_tutorial/ and definitely here: http://www.openmusiclabs.com/files/expotemp.pdf.
In brief, as far as I can tell, the only method used on the Synth pro is the CPU-based tuning scheme that utilizes digitally controlled potentiometers pulling the bases of each expo pair that connect to the pitch CV input towards ground. (I.e. no self-heating method identified, neither tempcos). Pitch CV per VCO goes directly to the expo pair from a CPU Pin via two 7k5 resistors in series, including a RC filtering cap in between them. As there is no method for “automatically” compensating pitch when the unit warms up (as a tempco would), hitting the tuning routine is the only thing we have (and, yet I’m guessing here, tuning tables do not seem to be saved in software as on some DSI synth, which “learn” by each calibration routine). Well, the synth works as it is, yet it would be sweet to improve tuning stability further.
One more “msytery” (to me, anyways) is that, according to a statement by Fricke, LM3046s are crucial to the VCO/DCO switching feature, and it would be super interesting to know how this works. [BTW, LM3046 were discontinued in 2017, yet Alfa and Coolaudio have remakes in the works or already available.]
As for waveform shaping and mixing, Waveform selection and final VCO mixing is done per digipot ICs (4315 – quad digitally controlled potentiometer). Per VCO you have one for blending between saw/tri and square/ringmod/noise and another for level control. Noise *seems* to come from the CPU.
VCO signals after the volume digipots are then summed via 100k resistors, pass a differential amp (need to check this one out more), and sent to VCF1 and VCF2 separately. [Two remaining digipots are for cutoff modulation and resonance duties on VCF2.]
One modification I just implemented and need to test in various setup is to increase VCO levels so that the filters have a bit more to chew on. Reducing the 100k summing resistors to 68k each sounds really nice so far! Clips soon…
CALIBRATION: should you ever need to calibrate filter tracking (for self-oscillation, sinewave patches etc.), here’s some post with information not found in the manual on how that’s done: https://www.sequencer.de/synthesizer/threads/mfb-synth-pro.150188/page-25#post-2204830.
VCF1 is a 12-db Sallen Key-style state variable filter built around a quad op-amp and a LM13700. The quad op-amp carries Lowpass on A (Pin1), Bandpass on B (PIN7), and highpass on C (PIN8). VCF2 is a SSI2144 lowpass. Voltage control is done via an Lm13700, and the resonance feedback path (which can be bumped up via software setting) includes diode clipping (see yellow box bottom right), which contributes to its gnarly characteristic. By the way, the input is attenuated by a 100k resistor, and reducing the value of that one might be an interesting modification 😉
Filter response selection (I think also parallel/serial setting) is done via an IC switch (HEF4053). Still need to do more tracing here, and need to understand where the stereo spread of VCF1 and VCF2 is done.
VCF2 is a SSI2144 low pass transitor ladder filter with the typical level drop at higher resonance settings. There’s not much in terms of periphery as far as I can see – frequency control, resonance control and qcomp is done at the digipot section.
VCAs and SUMMING
The VCA stage (actually two VCAs fed in different configurations dependoing on the streo spread setting) build around a LM13700. Control input per VCA comes from the CPU, passes a 3k6 resistor and RC filtering cap, then passes a common base NPN transistor, and finally another 3k6 resistor. All voices are summed right at the outputs of the VCAs of each voice and the signal then enters the final mixer and FX processor on the control board.
One modification I’m currently testing is to bump up the impact of VCA control a little. As is, the Synth pro is perfectly fine, yet I’m personally after a bit more bite. A great difference to me is to reduce the “upper” 3k6 resistors (i.e. those between CPU and PNP transistor) to around 2k6. Need to see how this agrees wih increased VCO level (another mod I’m just testing) and polyphony without resulting in clipping).
Some tentative Ideas for a single voice output (1+7 Mode)
This is still very tentative and needs more thought, but since the Synth Pro has different play modes, it’s also tempting. in 1+7 mode, one voice is allocated to midi channel 1 and the other seven voices to channel 2, so you can play bass + chords type of stuff (MKS-7 ahoy). That one voice is always the one marked on the picture below.
Unfortunately, separate outputs for each voice are not possible without cutting traces on the PCB. If you did that nevertheless for voice “one of eight”, which I cannot recommend, you could switch (using a dual switch) the two VCA outs of that voice to an alternative output. Best use a op-amp buffer at that alternative output (simple voltage repeater might be enough).