Discussion about the Korg PolySix synthesizer

Now I know how difficult tuning a polyphonic analog synth can be!

I have been through the tuning process a couple of times. Both times I hit the upper limit of tuning for most of the voices.

The problem is that there is no guideline in the service manual for
calibrating the antilog amp (before the multiplexer).
I always wanted to find out about this in more detail because I'm sure this
will affect the tuning precision over the whole range. I once had a Polysix
on my desk with an antilog amp that was very mis-calibrated. It took me
ages to get it into an acceptable tuning: The voices were mostly hitting
the designated tuning "spots" but were out of tune in between. Imagine a
diagram of a sine function - the tuning felt like that.
The Korg Trident manual has a procedure for this but it is not very
helpful (at least it didn't work for me).

This is also a circuit part that differs between the voice board revisions.

Perhaps it is possible to extract something from comparing it to the
service instructions of other synths from that Korg era. As I said there is
something about this in the Trident manual; and the MonoPoly manual also
has instructions (of course the circuit differs).

There must be some kind of procedure; I can't imagine they did this by
trial and error at the factory.

Do you think this would this explain why the VR is acting weirdly? Why does the pitch rise after a while after I have turned the VR anti-clockwise? It does exactly the same thing in the other direction.

At 6:03 Uhr -0800 28.02.2014, <josh.nursing@gmail.com> wrote:

> Do you think this would this explain why the VR is acting weirdly? Why does
> the pitch rise after a while after I have turned the VR anti-clockwise? It
> does exactly the same thing in the other direction.

I think this is more or less normal behaviour. There are three VR for the
calibration of the antilog amp. All three interact somewhat. I do remember
that the behaviour you describe was similar to some setting on my machines.

If I find the time I'll look into this soon.

Thanks Bob, Malte and Dennis, I will check these and try again.

Does anyone have access to a complete (!) Trident Mk II service manual?
Posting the procedure for the antilog calibration would help tremendously.

I have the calibration routines for:

Trident (Mk I) -> incorrect and partly implausible
Poly-61 -> will have to check
MonoPoly -> will have to check


Can anyone think of other Korg Polysynths from that era that might use the
same antilog technique? I think this is the list:

Polysix
Poly-61
MonoPoly
Trident
Trident Mk II

It would seem odd to me that all of them have an antilog calibration
routine except the Polysix. The Polysix also has a test point at the same
circuit position...

Hi,

this reminds me of something.
I believe at least the Polysix and the MonoPoly use the same assigner.

Polysix 8049C217
Monopoly 8049C217

Same for the old revision of the Poly-61.

The Poly-61 (being a 6-voice synth) could lend useful insights here.

The MonoPoly has a "Tune Point Switch" that - as far as I can see -
connects pin 34 to pin 19 for the initial tuning procedure.

> The voltage at TP2, called Oct/V CV, which seems to be the input
> to the antilog ckt, is constantly switching between 8 values. Six of them
> are the same when you go into unison mode, but there are still two other
> ones being muxed in. I guess for the factory calibration they must
>have had
> a way to pause the muxing and put out specific DAC values for setting the
> trimmers using a DMM. Seems like that code would be in the keyboard
> MCU, unless they had a factory test MCU that they used.

At 14:27 Uhr -0800 28.02.2014, Bob Grieb wrote:

> Found this on the web:
>
> "The other test mode is started by jumpering pins DB7 and P17 using a
>diode (cathode towards DB7) on the
> 8049 processor on the KLM-366 circuit board (this can be done on the
>keyboard connector CN04). In this mode the
> CV DAC is set to generate 0V, 2.5V, or 5V depending on the position of
>the OCTAVE switch (16' results in 0V, etc.).
> This mode is used to check the CV DAC and is not described in the service
>manual."
>
>
> I wonder if this could be used to set up the antilog circuit? Or
>possibly once in this mode, there are
> other voltage settings as well? Think I will try this out.

That's exactly what the MonoPoly does!

The Monopoly service manual however gives

0.000V Down Low Adjust
5.250V Normal Mid Adjust
10.583V Up Hi Adjust

and takes the signal after the demultiplexer (and after the buffer and
possible a 100R resistor).

The octave switch in the Polysix should theoretically do the job.

Yes, it's possible that the circuit in the MonoPoly isn't an antilog
circuit; but still the calibration involves some scaling. It's apparently
not completely linear, is it?

As far as the three voltages go: This is exactly what the other antilog
calibrations do, too.
The Trident requires to press the same key 8 times to load the DAC with the
same voltage. This is done for three different keys (C1, C3, C6).
The Poly-61 uses unison mode but measures just for single voice (apparently
after the demultiplexer).
So my guess is that three voltages are indeed enough.

If I'm not mistaken the antilog circuit for the Polysix has four trimpots
(at least in the new version): one each for low, mid high and another one
for a general offset. I think the old version only had three pots.


At 15:08 Uhr -0800 28.02.2014, Bob Grieb wrote:

> Yes, it seems we can get three different static DAC values, but I am not
>sure that
> is enough to calibrate the anti-log circuit. I tried the test, and saw
>roughly 0,2.5 and 5V
> on the output of the DAC. I pressed lots of front panel switches.
>Other ones also affect the voltage, but it seems there are only three
>choices. Seems like you would want more voltages, and specific ones that
>were chosen for the job. The circuit in the Monopoly that is being
>adjusted is not logarithmic, it seems. Just a gain stage.

Hi,

There is some Trident MkII stuff in the files section in this group. Including some tuning info.

Kind regards,
Ben

On 28 feb. 2014, at 21:54, Malte Rogacki <gacki@gacki.sax.de> wrote:

> Does anyone have access to a complete (!) Trident Mk II service manual?
> Posting the procedure for the antilog calibration would help tremendously.
>
> I have the calibration routines for:
>
> Trident (Mk I) -> incorrect and partly implausible
> Poly-61 -> will have to check
> MonoPoly -> will have to check
>
>
> Can anyone think of other Korg Polysynths from that era that might use the
> same antilog technique? I think this is the list:
>
> Polysix
> Poly-61
> MonoPoly
> Trident
> Trident Mk II
>
> It would seem odd to me that all of them have an antilog calibration
> routine except the Polysix. The Polysix also has a test point at the same
> circuit position...
>
>
> ------------------------------------
>
> Yahoo Groups Links
>
>
>

Thanks. This is similar to what I did to get the mis-calibrated Polysix
back in tune.
There are two problems with this though:

1. If the calibration of the voice is off the mark it won't work reliably.
2. If the calibration of the "middle" antilog is off the mark it won't work
reliably.


And this can be a big issue: If the antilog wasn't properly calibrated to
begin with it's quite likely that someone will have tried to tune the
voices but hit extreme positions and so on.

The calibration procedures for the Trident Mk I, the Polysix and the
MonoPoly give definite voltages for which to calibrate. This has the
advantage that afterwards you have a perfectly calibrated voltage source.

At 1:10 Uhr +0100 01.03.2014, Ben Stuyts wrote:

> There is some Trident MkII stuff in the files section in this group.
>Including
> some tuning info.

The Mono/Poly uses a completely different VCO design.  The Mono/Poly has Volt/Octave VCOs instead of the Volt/Hz found in the Polysix.  So, the tuning issues are very different.

While indeed the oscillator has a different design it should still be
possible to get the antilog amp calibrated better.

I actually think that the whole test mode isn't even needed: What we
actually need is the defined voltage coming from the DAC and then the
target voltage for a single note after the demultiplexer.
This seems to be the procedure for the Poly-61 (it uses unison mode so that
all voices have the correct voltage). This should nicely circumvent the
problem that there are two unused voices with the 8049C217.

The Trident measures the voltage for all 8 voices together in front of the
demultiplexer; hence it is necessary to press a single key eight times.

However I don't know if the antilog amp somehow has to account for
non-linearities in the oscillators.

I'll give this some more though later.


At 6:45 Uhr -0800 01.03.2014, <chipaudette@yahoo.com> wrote:

> The Mono/Poly uses a completely different VCO design. The Mono/Poly has
> Volt/Octave VCOs instead of the Volt/Hz found in the Polysix. So, the tuning
> issues are very different.
>
> As a result, I don't think that the Mono/Poly will be much help for
>insight on
> how to recover the tuning on a Polysix.

Another idea:

They built in three TPs that are only useful for calibrating the antilog amp.
TP4 first caught my eye. This is -10V.

With the Trident you were required to attach the multimeter ground point to
a -11V point.

All the following measurements are done on a new production voice board.

I checked what happens when connecting P17 to DB7 with a diode.

The voltage at TP2 went

16': -8.62
8': -11.22
4': -13.33

Now, TP2 is conveniently labelled "Oct/V CV" while TP3 is labelled "Hz/V CV".

Voltage at TP3 is:

16': -9.94
8': -8.75
4': +4.20

Now, if we connect the multimeter ground probe to the aforementioned TP4 we
get (more or less as expected):

16': +0.03
8': +1.23
4': +14.19

That this doesn't up a 100% with the TP3 voltages may have something to do
with the voltage on TP4 not being 10.00V.

Anyway; I think somewhere around this must be the crucial point for the
calibration.

It should also be noted that the voltage at the "OCT/V CV" TP2 is NOT
1V/Oct. It's pretty precisely 0.5V/Oct. This is also visible in the service
manual on the page for the old production key assigner.

I have the feeling we're getting closer.

Perhaps the factory used a "dummy plug" with the diode connecting P17 and
DB7 instead of the keyboard for calibrating the antilog amp. Given the fact
that they also had another dummy plug with the 56k resistor for calibrating
the reset circuit I find this not unlikely.

Measurements on and old production board. TP2 and TP3 are obscured by the
KLM396A board which must be removed.

TP2

16': -8.40
8': -11.02
4': -13.30


TP3

16': -9.93
8': -9.39
4': -2.40

TP3 with ground probe at TP4

16': +0.03
8': +0.97 (somewhat unstable occasionally more around +0.60)
4': +14.20

The last pair is a a bit of a mystery to me. Perhaps my multimeter is
freaking out.

And one final thought before I retire for tonight:

I don't think the test mode for the key assigner puts out 0, 2.5 and 5V
respectively.

The test mode for the MonoPoly put out a much larger voltage than the whole
keyboard required. And it was not directly related to the CV.

Monopoly:

0V
5.25V
10.583V

Polysix new:

-8.62
-11.22
-13.33

Polysix old:

-8.40V
-11.02V
-13.30V


This also did not correspond with the regular key voltages (0.5V/Oct). If I
try to relate the numbers by removing the offset I get for the Polysix:

Polysix new:

0V
2.6V
4.71V

Polysix old:

0V
2.62V
4.9V

It's obvious that the two top values don't fit into this.

For the Monopoly the ratio for middle value to top value was 1 to 2.015.
For the new Polysix it was 1 to 1.811.
For the old Polysix it was 1 to 1.870.

Anyway, since for both the Monopoly and the Polysix the values are somewhat
odd I could imagine that those values were picked to simplify the antilog
calibration. However in the Monopoly this makes not much sense (you have to
calibrate to those three values to get regular 1V/OCT response - why not
use values like 5V or 10V then?) I could imagine that this is a Polysix
"thing".

And here's the voltage for all "C" keys, starting from the lowest C and 16'
to the highest and 4'. Measured on the new production board, after the
demultiplexer. Stretch tuning is off (this is important!). Multimeter
ground is TP4 (-10V).

0.018
0.042
0.090
0.185
0.375
0.755
1.513
3.03x

The doubling of the voltage for each octave is easy to see. I think this
antilog amp has a good calibration; perhaps at the bottom there is a little
bit of room for improvement.

So we have essentially a doubling for every 0.5V at the input.


With VR1 to VR3 we can change the relation of those voltages for certain areas.

The service manual says:

VR1 ADJ. CENTER
VR2 TUNE HIGH
VR3 TUNE LOW
VR15 TUNE MID

However I'm not a 100% convinced that the functionality of VR15 and VR1
isn't switched.

VR15 seems to control a general offset while the other VR's indeed control
their respective ranges. They're affecting the tuning a bit outside their
range but not much.

A comparison of the old and new production also hints at this:

VR1 MID
VR3 LO
There seems to be no VR2 for the old production.

And the addendum says

"VR15 on KLM-366 is equal to VR1 on KLM-396."
The latter is the daughterboard for the old production; hence VR15 is NOT
the same as VR1 (which is simply labelled "MID") on the old production
KLM-366.

So my guess is that the four VR's for the new production should better be
labelled:

VR1 TUNE MID
VR2 TUNE HIGH
VR3 TUNE LOW
VR15 ADJ. CENTER

I've managed to improve the scaling a bit.

Before:

0.018
0.042
0.090
0.185
0.375
0.755
1.513
3.03x

After:

0.023
0.045
0.090
0.182
0.366
0.740
1.490
2.98x

VR1 has the least effect; that's the one that has the tuning "drift back" a
bit after adjusting it. But it doesn't go back completely.

Admittedly VR2 is now in its extreme position. There probably is still room
for improvement. It would be nice to have a definite table for all this.

I've done all the measuring with the "Tune" knob in the middle and the
pitch wheel centered. This voltage is also part of the calibration;so if
you have the tuning knob in a different position you'll end up with
different values.

And finally:

I've tuned the Polysix (so far just preliminary). There still is some small
room for improvement but except for the lowest 16' octave I have a pitch
deviation better than the one shown in the service manual. Some notes have
a 1 to 2 cent deviation but most are spot on. The lowest notes in the
lowest octave have currently a deviation of about 7 cent.

I had to use VR15 a lot though which means that probably the basic offset
was not so good. The whole antilog circuit calibration could probably be
shifted a bit upwards and then compensated for again with VR15. This could
possibly bring VR2 away from its end position.

This is just from the top of my head:

The optocoupler is in a feedback loop from the demultiplexer output to the
antilog amp.
If I remember correctly there are two unused "slots" in the DAC (and hence
two "unused" multiplexer outputs). But those two "slots" are always at the
same fixed value (I think). By using such a feedback loop it should be
possible to account for any changes to the circuitry (temperature,
voltage...) because this output would carry the changes and hence counter
them in the optocoupler.
I haven't checked what exactly the second "unused" multiplexer output does.
That's the one that leads back to vicinity of the 4051 in front of the
antilog amp (or the one on KLM-396 for old production boards).

At 6:15 Uhr -0800 02.03.2014, Bob Grieb wrote:

> One thing I don't understand is how the 4051's are being used to change the
> behavior of the circuit. And what is the fcn of the opto-coupler? Any
> thoughts?

That schematic is very nice and very helpful. That particular portion of
the service manual from the files section isn't the most readable.

One thing that immediately becomes clear to me now is that IC19 generates
the -10V rail. R26 (at the right) would also connect directly to this rail
but perhaps it's easier to read the way it is now.

The multiplexer/demultiplexer thing:

Correction voltage from VR1 enters IC39 X7. This voltage must come back
from IC31 X7 through IC18 at the bottom left of your schematic.

The two controlling signals for this seem to be MUXSELA for the Offset
correction (X7) and INH for the other correction (X6 to optocoupler).

I wrote my last reply, sent it, checked my mail and there was your
description of the same circuit, just much better.

:-)

After so many years I bid you wonderful Poly 6 people adieu. This is a great resource and a benefit to all Poly 6 owners.

Hi John,

Thanks for being a member of the group for all these years. Hope you are well, and take care!

Ben (Moderator)

On 2 mrt. 2014, at 20:23, john_terhorst@yahoo.com wrote:

>
>
> After so many years I bid you wonderful Poly 6 people adieu. This is a great resource and a benefit to all Poly 6 owners.
>
>

Very interesting: it is when I adjusted VR1 again while attempting tuning with a chromatic tuner that I encountered the weird behaviour described previously: on turning it clockwise, the pitch does rise, but when I reach maximum, it starts falling again. The similar odd behaviour happens in the other direction.

This behaviour is somewhat normal.

Please try the following:

Put the Polysix in manual mode. Tuning knob centered, pitch wheel centered,
no LFO modulation. Switch to Unison mode.
Please measure the voltage at IC31, pin14 for each "C" from bottom 16' to
top 4'and post that here.

I forgot: Please put the ground connection of the multimeter at TP4 (-10V).

And stretch tuning needs to be off - put the switch away from the connector.

Thanks for that, Malte. The readings yielded (my voltmeter is off by +0.02 so 0.03 should be read as 0.01):

To use only one expo, Korg has used multiplexers to feed the 6 VCO CVs to the expo, followed by a calibration voltage and then the CV for standard pitch (only for the new production voice board).
The two latter outputs are routed to two independent controllers; the last one introduces an offset to the expo input to shift the standard pitch to its desired value, while the converted calibration voltage drives a current through the photocoupler’s LED, while its LDR is placed where you would normally expect the 3300ppm tempco resistor in similar circuits.

This leads to an eady way to check whether a tuning problem is related to the photocoupler circuit at all: measure the voltage on pin 7 of IC18. Within regulation it should be somewhat between -2 and -6 volts, -3 are more normal.
When the loop breaks open due to failure of the photocoupler or related parts, it will be stuck on the maximum negative output, something around -13 volts. There must always be a small current through the LED, otherwise the loop does not work and the TUNE HIGH preset has no effect!"

2. From Chip's blog: http://synthhacker.blogspot.ca/2013/02/polysix-basic-key-assigner-timing.html

"Looking at this graphs, I see that overall timing loop is 6204 microseconds ("usec") long.  This means that the basic pitch and the on/off gate of each voice only get updated every 6.2 milliseconds, which is an update rate of only 161 Hz. Frankly, I was surprised at how slow this is.  But, since the synth sounds OK, I guess that it works fine.  Luckily, the pitch modulations and the VCF and VCA envelopes are not at all dependent upon this loop (they have their own generators which, for the case of the VCF and VCA, are fast).

Additional examination of these graphs (and a little more probing of the synth) allows one to understand the timing cycle well enough to break it into discrete time periods.  Clearly, there are six periods associated with the pitch of the six voices.  I call these periods "V1" through "V6".  These periods are each 676 usec in duration.

After V6, there are appears to be another voice-like period, which I call "Vx7", since it is so much like a voice, but doesn't actually sound.  Interestingly, through additional experiements, I found that its voltage is always the same - it is always set to be C7, the highest note on the Polysix.  Vx7 is 712 usec long.  Why is it a different length than V1-V6?  I don't know.

Similarly, after Vx7, there appears to be another voice-like period, which I call "Vx8".  Its voltage level always appears to be identical to V1.  I don't know why.  I'm thinking that Vx7 and Vx8 are both used by the Polysix's pitch correction circuitry to help it stay in tune.   The duration of Vx8 is only 624 usec (though the end of Vx8 is actually a bit hard to define and measure).

After Vx8 is the final period of the Key Assigner's timing cycle.  I call it the "Inter-voice" (or "IV") period.  During this period, the synth sets the gate signals, it drives some LEDs, it scans the keybed and some switches, and probably a few other things as well.  We'll dig into that in a future post."

Yash

I think you need a more precise multimeter for this. Right from the reading
I'd say the scaling is slightly off and the tuning could be slightly flat
toward the top. You'll notice that the doubling usually ends with one
number lower than expected.

I've just did another test - I intentionally miscalibrated both antilog amp
and the voices. Within about 20 minutes I was able to get the tuning back
within the standard deviation chart. A better tuning might take a bit
longer.

I found the target voltages a little bit higher than your measurements.

I started out with the VR1 to VR3 and VR15 roughly centered.

Then I tried to get the highest note of the 16' setting close to 0.750 and
then worked from there.

I currently have the voltages as follows:

0.023
0.046
0.093
0.186
0.375
0.750
1.502
3.00

This results in a comparatively satisfactory tuning with no tuning VR in
its extreme position.

I initially used VR15 to get close to the 0.750 value. VR1 is only useful
for very small finetunings; I think it affects the 0.750 setting (which I
consider roughly "the middle" about +/- 2mV.

For tuning I use the following procedure:

I start with the first voice and tune 4' C6 as described in the manual.
Then I press C2 six times so I'm back at the first voice and adjust VR10 as
described in the manual. Without releasing the key I switch to 16' and
adjust according to the manual. This will usually again affect the previous
setting so I'll switch back to 4', make adjustments, switch back to 16',
make adjustments and so on. It usually takes about ten to 20 steps until
I'm close.
After this I'll usually recheck 4' C6 for that voice, readjust and repeat
the procedure.
Once the first voice hits all relevant notes with 1 cent or less deviation
I'm tuning the remaining voices.
This is now much faster because I don't have to worry about the VR15
setting for the remaining voices. So it's just top and bottom.


If I was in your place I'd probably try to turn VR15 slightly to the left
to increase all voltages a bit.

Another mistake again:

At 23:25 Uhr +0100 03.03.2014, Malte Rogacki wrote:

> I start with the first voice and tune 4' C6 as described in the manual.
> Then I press C2 six times so I'm back at the first voice and adjust VR10 as
> described in the manual.

This should read "and adjust VR15 as described in the manual". This is step
4 of the tuning instructions. In the service manual this describes KLM-396
VR1 (the VR on the small daughterboard); this is the same as VR15 for the
new production voiceboards.

Malte, thanks for following up and sharing your settings.

I will have to check if I haven't mistaken VR1 for VR15.

At 14:36 Uhr -0800 03.03.2014, <josh.nursing@gmail.com> wrote:

> I will have to check if I haven't mistaken VR1 for VR15.
>
> My board is a new production one.

VR1, VR2 and VR3 are next to the strech tuning switch.

VR15 is directly left from the first voice on the new production board.

For new production boards step 4) if the tuning instructions should read:

"Play C2 and adjust VR15 to obtain 0 cent. Do this for one unit."

There is a pretty good chance that following this instruction will already
bring you back into a more friendly territory tuning-wise.

I managed to get very good tuning both to my ears and thanks to a Mac OS chromatic tuner.