The MS20/Trident discussion from the begining!!!

[Bjorn Julin]

Hi,

Since the list was "DOWN" all my posted emails and probably
the emails from Juergen newer showed up and i had put a lot
of efforts writing them and i got realy pissed off by the
list server machine was rejecting my fucking emails, so now
i send them all once more!!!!!! It might be parts of the
discussion that has some interest to some of you!

If any has some knowelege about any of the questions
1 to 11 i really aprechiate the possible answers.
Or any of the other questions in the later emails.

I started the discussion in what i believed was that the Trident
had linear detuning, and so did juergen for a while until it was
discovered by juergen that it had "not" linear detuning by "default".

So to the folks that are interested in linear detuning and Korg CCO's
and newer got mine and Juergen's discussion kan have it once again.

Reg
BJ
---------------------------------------------------------------------

Hi folks! And Juergen!

This might be old and "already discussed" on the
diy list if so please point me to the year of
discussion held so i then can look in the archive.

Otherwise, What a brilliant design!!!

The more i look into the deign of Polysix and Trident
the more fascinated i get.Its so clever, just take
look at the oscillator core three transistors,one cap
one resistor and a diode and it performs absolutely
brilliant! With a ref current of 6,8uA  and a charge
cap of 4,7n it have a frequency range of 62Khz to 0,0074Hz!!!

I cant even measure it with a frequency counter, nor a scope,
i had to sit with a stop watch and count the seconds dragging
away!! I used just standard stuff like BC557,BC547 and a LF353
as JFET buffer.

I think its possible to achieve even lower frequencies,
i don't think the core is the problem, its the expo converter
that have difficult to control the small currents of such
low frequencies. Also the temp drift of the OP amps in the
expo start to hassle and stress.

It seams to be quite linear to, i haven't measured yet but
it changes its amplitude very little at the very far end of
the frequency spectrum, Has anyone any experience of the
Korg oscillators linearity? Jurgen?

Another thing that's puzzles me are that in the Trident and
Polysix they just bypass each OSC with one 100n cap!!!
The osc must be very Soft sync friendly, and that in a
16 VCO machine!!! Any ordinary LM393/311 or CEM3340
based machine would soft sync like hell whit the similar
bypass!!I couldn't even get my three VCO MAS board (CEM3340)
to stop soft sync until i had some gnd plane and very
carefully laid PCB.

And just look at the temp compensating stuff for the Polysix,
a vactrol adjusts the resistor division ration of the CV voltage!!
,also since the vactrol compensation voltage are one of the mux
channels the cirquit compensates each CV channels Ron value and
any temp variation that might occur in the 4051 mux. Not entirely
perfect, but Brilliant, just damn brilliant!

It seams that there was a old and new version of the Polysix
the old had a discrete DAC and no mid tune cirquit but the same
vactrol compensation, the Trident on the other hand used
the same expo converter as the Old Polysix but no vactrol
but instead a heated uA726 and a resistor called TH1 in
series with the CV voltage and parallel with a series resistor.

The third interesting thing is linear detuning, Juergen has talked
a lot about how to create linear detuning in exponential VCO's in
an old email he was talking about different ways and one of them was:

>"Another way to introduce an offset is using an exponential 
>voltage-to-voltage converter (just drive the collector current
<into a resistor instead of the VCO capacitor), and use this
<voltage to drive a linear VCO. Some Korg Synths did this
<(Polysix, Trident), and as far as I know the first Moog
<Modular VCOs as well."

This probably became Magnus expo to Hz/v converter, never the
less one thing Juergen might have missed was that the Trident
in fact ""have"" linear detuning!!!

Im pretty shure about Juergen did a original invention and im quite
convinced that the Korg engineers didn't know that it was linear
detune, they had implemented!!! Anyway Juergen shall have all the
credits for this implementation on linear detune, if he had not
been talking about it i would never had been searching for
it in the first place.

Back to the Trident, what we have missed are that in the
design of the Trident they have created a controllable
offset voltage to the charge transistor, by pushing the
thresh hold level more and more positive we charge the
cap sooner and sooner, linear detune folks!

In the case of the Korg engineers, they where probably only
looking for a simple way (low component count) to have a
detune function.If they had chosen to implement it before
the exponential convertor they would have needed quite a lot
of CV's and multiplexers.

But on the other hand, they did implement the string
section in the Trident!!! That is quite creepy stuff
folks, how much did the Korg engineers actually know??

As juergen once said:
"Most stuff has already been designed"",

BJ says:
Well, one never knows! :-)

My questions:

1:Are the TH1 resistor a NTC for temp compensating in the Trident?
If so, why, as compensating temp drift in the 4558 OP
in the expo cirquit? Or what?

2:How temp stabile are the MS20 vco core?
Any measuring anyone?

3:Playing a Polysix in unison how much beats
the VCO's against each other?

4:Has anyone seen any Korg patent of linear detune?

5:Did Korg patented the MS20/Polysix VCO core?

6:Did Korg patented the muxed expo converter?

7:The principles of the MS20 VCO, and the Moog VCO
are almost the same, one can tie the charge cap
to GND as well as to +V, but are there any significant
advantages by tying it to +V?

8:Which one came first, the Polysix or the Trident?
Wen did the Trident MK2 came?

9:does anyone have any hands one with a Trident or Trident MK2,
how does the tuning accuracy feel when runned parallel with a
digital machine, on tune or of tune?

10:Anyone having a url for specs over the SC and SA transistors?
(have already looked around at Matsushita and Rhom etc).

11:Has anyone listen real close to how close the Trident
osc's can beat?

Finally,i took a look at Juergens poly stuff and it looks interesting,
maybe i go poly too, if so i go the Trident way, the big hassle for
me is the poly trigger keyboard interface! It seams that BJ has to
learn to program a PIC device! Urgh, kvack!!!

Regards
BJ
-----------------------------------------------------------------------------------

>    glad to see more people valuing the brilliance of Korg
<    designs. ;-) Yes, that's definitely something to rave about.

Yep!

Besides one can put the charge cap to gnd as well.
you get a slight increase in amplitude in the case of MS20.

>    In fact, they needed linear *tuning* to get
<    rid of the offset voltage of the opamps they used.

Wich one do you refering to?

In the MS20 they used a trimmer on the TL071 to trim out the
offset voltage! In the Polysix and Trident they trim any
errors in the muxed expo, the other trimmers at the
drain side of the current regulator seams to be only
zero adjust and fine adjust of high freq!

Besides how does a 1:1 current regulator performs? (Psix/Trid)
Is there any severe temperature or linearity dependencies
in the transistor of the regulator? Or in that design in
regular?

-----------------------------------------------------------------------------
Hi Juergen and all!
And thanks for your reply, i realy aprechiate it!:-)

>>With a ref current of 6,8uA  and a charge cap of 4,7n it have a frequency 
>>range of 62Khz to 0,0074Hz!!!

>Yes, you can actually tune the offset voltage of the opamps by ear: With no 
>input CV applied, turn the pot until the tone disappears and
< you get no click for a minute ...
< Or adjust linear detuning: with no CV applied, adjust offset until
< you hear it clicking at the desired beat rate.

You mean the OP's in the expo converter?

>>I cant even measure it with a frequency counter, nor a scope,
<> i had to sit with a stop watch and count the seconds dragging
<> away!! I used just standard stuff like BC557,BC547 and a LF353
>>as JFET buffer.

>I guess I'll try some of my standard transistors as well then. So far I've 
>used the japanes transistors from the Korg schemos.
<I can imagine leakage currents are an issue here, and the discharge
>time will depend of the transistor type as well ...

Me to,but what else besides high beta and low leakage,
would low Re be good in the core? I just see the transistors
function as a current pump and a treshold switch and since
they are the core they are quite fast and therefor they would
be quite linear? or am i wrong into this?

>>It seams to be quite linear to, i haven't measured yet but >
>>it changes its amplitude very little at the very far end of the frequency 
>>spectrum, Has anyone any experience of the Korg oscillators linearity? 
>>Jurgen?

>No measurements from my side. I'm only using it over
<a couple of octaves anyway; the rest is done with dividers
<(as in Minikorg). But in the MS-20 it's full range, and I've
<never heard anybody complain about the MS-20's linearity.

The trident uses sums of saw and square dividers to get octaves
but the polysix does not, the Psix has a range of 7 octaves
5 by KBD and two octaves (as it seams to me, but not 100%)
switched by a 4051 to the base of the muxed expo koverter.

I have a idea, what about multiplexing the ref current instead
to get octaves?

>>Another thing that's puzzles me are that in the Trident and
<> Polysix they just bypass each OSC with one 100n cap!!!
<> The osc must be very Soft sync friendly, and that in a
<> 16 VCO machine!!!

>I'm reluctant to believe that it's less critical than other
>designs. Maybe the single cap is combined with a clever layout
>that works like a small inductance ? (I've never seen the
>Trident layout - just guessing.)

Trident is mor or less the P six in the core, the difference
between MS10-20-50, Trident and P six are the transistors used
in the core.

Newertheless i tested two cores on a lab board, plastic
bread board with only one 47u elyt decoupling and wires all
over the place and "NO soft sync" these cores must be good!

>>And just look at the temp compensating stuff for the Polysix, a vactrol 
>>adjusts the resistor division ration of the CV voltage!!
<> ,also since the vactrol compensation voltage are one of the mux
<> channels the cirquit compensates each CV channels Ron value and
<> any temp variation that might occur in the 4051 mux. Not entirely
<> perfect, but Brilliant, just damn brilliant!

>Truely brilliant. And rather perfect in a practical sense as well:
>The expo converter's temperature drift is fully compensated, but the
<(partly desirable) linear drift of the buffers after the expo conv
<is not.

So what you are saying is a positive temp drift in the
buffers compensates for high frequency flattness in the
core?

>>the old had a discrete DAC and no mid tune cirquit but the same vactrol 
>>compensation, the Trident on the other hand used the same expo converter 
>>as the Old Polysix but no vactrol
<> but instead a heated uA726 and a resistor called TH1 in
>>series with the CV voltage and parallel with a series resistor.

>There is one drawback of that method, of course: It's hard to perform expo 
>modulation on VCOs with shared expo converters. The Trident is quite 
>lacking here.

But the Psix and Trident uses the same modulation cirquit,
multiplex and sum the MG and CV for each cannel and then
feed it to the expo converter.Ifcat i find it quite praticall
in apolly synt , for example multi plex 6 CV to the rate
and amplitude of the MG generator and you get 6 independent
Lfos.

>>Back to the Trident, what we have missed are that in the design of the 
>>Trident they have created a controllable offset voltage to the charge 
>>transistor, by pushing the thresh hold level more and more positive we 
>>charge the cap sooner and sooner, linear detune folks!

>>In the case of the Korg engineers, they where probably only
<>looking for a simple way (low component count) to have a
<>detune function.If they had chosen to implement it before
<>the exponential convertor they would have needed quite a lot
<>of CV's and multiplexers.

>You're right - I never noticed that. bass sound were due to remaining 
>tuning errors (offset
>voltage drift),but looking at the schemos again: yes, right, they do their 
>intentional controlled detuning
>as you described. (And with a lot of bypass caps here, too.)

>to sound quite boring in the upper registers. I always
>thought the linear detuning feature an *addition*,
>not a replacement, for expo detuning.

It seams that it has only lin detune, exept for the expo LFO
modulation, but i find it to sound quite interesting,a constant
beat over the entire range sounds like well defined strings!!
As and old Solina string machine.I think that in the case of
Trident they just wanted to do detune as cheap and quick
as possible.But i think that "no one" did know that it
infact was lin detune they had implemented in the Trident.

If you newer had invented it "again" i would for sure
would newer had reflected over it in the Trident.

---------------------------------------------------------------

>>Just like the Moog Taurus: linear detuning due to offset
>>voltage drift,plus a Detune pot that works the ordinary (expo) way. (I 
>>only have a copy of the VCO page).

Some one on the list had it uploaded onto the net!!

This is the interesting part, when you say it, it comes to
my mind that there are infact another machine that has
linear detune.

The Bit99 (crumar).The Bit uses 2 VCOs as master osc's the
oscs are multi vibrators who are driven by two separate
current regulators (no expo function) and a pot called
detune are controlling one of the current regulators.
And that is definitevely linear detune. The frequencies
are then driving 6 8253 counters used note dividing.

But the most interesting part is that the machines that are
best on tight bass sounds are the one with linear detune.
I have heard the Taurus and the bass is  really bommy and
so are the Bit99/01 and so would the Trident
be to if the Oscs are close in the bass range.

Allso all the machines that implements lin detune will have
good string  sound impelementations. it was a long time ago
when i heard a Trident but i can recall it to have a good
string sound (synth section, not string section) The bit99
has exelent string sounds. And a polly Taurus would have it to!

Reg
BJ
----------------------------------------------------------------
>From: jhaible at t-online.de (jh.)
>To: "Bjorn Julin" <bnillson at hotmail.com>
>CC: <synth-diy at node12b53.a2000.nl>
>Subject: Re(2): Lin detune,MS20 Core,Polysix/Trident
>Date: Sun, 15 Oct 2000 13:52:38 +0200
>
> > With a ref current of 6,8uA  and a charge
> > cap of 4,7n it have a frequency range of 62Khz to 0,0074Hz!!!
>
>Yes, you can actually tune the offset voltage of the opamps by ear:
>With no input CV applied, turn the pot until the tone disappears and
>you get no click for a minute ...
>Or adjust linear detuning: with no CV applied, adjust offset until
>you hear it clicking at the desired beat rate.
>
> > I cant even measure it with a frequency counter, nor a scope,
> > i had to sit with a stop watch and count the seconds dragging
> > away!! I used just standard stuff like BC557,BC547 and a LF353
> > as JFET buffer.
>
>I guess I'll try some of my standard transistors as well then. So
>far I've used the japanes transistors from the Korg schemos.
>I can imagine leakage currents are an issue here, and the discharge
>time will depend of the transistor type as well ...
>
> > It seams to be quite linear to, i haven't measured yet but
> > it changes its amplitude very little at the very far end of
> > the frequency spectrum, Has anyone any experience of the
> > Korg oscillators linearity? Jurgen?
>
>No measurements from my side. I'm only using it over
>a couple of octaves anyway; the rest is done with dividers
>(as in Minikorg). But in the MS-20 it's full range, and I've
>never heard anybody complain about the MS-20's linearity.
>
>
> > Another thing that's puzzles me are that in the Trident and
> > Polysix they just bypass each OSC with one 100n cap!!!
> > The osc must be very Soft sync friendly, and that in a
> > 16 VCO machine!!!
>
>I'm reluctant to believe that it's less critical than other designs.
>Maybe the single cap is combined with a clever layout that
>works like a small inductance ? (I've never seen the Trident
>layout - just guessing.)
>
> > And just look at the temp compensating stuff for the Polysix,
> > a vactrol adjusts the resistor division ration of the CV voltage!!
> > ,also since the vactrol compensation voltage are one of the mux
> > channels the cirquit compensates each CV channels Ron value and
> > any temp variation that might occur in the 4051 mux. Not entirely
> > perfect, but Brilliant, just damn brilliant!
>
>Truely brilliant. And rather perfect in a practical sense as well: The expo 
>converter's temperature drift is fully compensated, but the
>(partly desirable) linear drift of the buffers after the expo conv
>is not.
>
> > It seams that there was a old and new version of the Polysix
> > the old had a discrete DAC and no mid tune cirquit but the same
> > vactrol compensation, the Trident on the other hand used
> > the same expo converter as the Old Polysix but no vactrol
> > but instead a heated uA726 and a resistor called TH1 in
> > series with the CV voltage and parallel with a series resistor.
>
>There is one drawback of that method, of course: It's hard to
>perform expo modulation on VCOs with shared expo converters.
>The Trident is quite lacking here.
>
> > Back to the Trident, what we have missed are that in the
> > design of the Trident they have created a controllable
> > offset voltage to the charge transistor, by pushing the
> > thresh hold level more and more positive we charge the
> > cap sooner and sooner, linear detune folks!
> >
> > In the case of the Korg engineers, they where probably only
> > looking for a simple way (low component count) to have a
> > detune function.If they had chosen to implement it before
> > the exponential convertor they would have needed quite a lot
> > of CV's and multiplexers.
>
>You're right - I never noticed that.
>I thought the linear detuning effect that gives the Trident its gorgeous
>bass sound were due to remaining tuning errors (offset voltage drift),
>but looking at the schemos again: yes, right, they do their intentional
>controlled detuning as you described. (And with a lot of bypass caps
>here, too.)
>Does this mean the Trident has *only* linear detuning ? Then I'd expect
>it to sound quite boring in the upper registers. I always thought the 
>linear
>detuning feature an *addition*, not a replacement, for expo detuning.
>Just like the Moog Taurus: linear detuning due to offset voltage drift,
>plus a Detune pot that works the ordinary (expo) way.
>It's been a long time since I heard a Trident, so please correct me if
>I'm wrong here. I think it's also time for me to get the full service
>manual (I only have a copy of the VCO page).

> > But on the other hand, they did implement the string
> > section in the Trident!!!

>And that wonderful trigger selector for the Brass section.
>(This comes from the PS-series, btw.)

>JH.

----------------------------------------------------------------------
Hi JH and ALL.

>>I dont agree, when i look at a Korg CCO i see a core
>>wery similare to a Moog Rouge etc, core for example.
<> The principle of "charge full" instead of the usuall
<> short the capp is wery similare. But you are saying VCO
<> and that is confusing!
>[...]
>>Still i dont agree in this reasoning about what is expo and lin
>>CCO's to me they are all lin CCO cores but with different principles
>>in the front end of the Core.I find it quite confusing when
>>people refer to it as a expo or lin VCO, i think this issue has
>< to be sorted out once for all!?

>... and I had thought I was the one who helped sorting it out on
>this mailing list (;->).

Yes you have, the problem is that i have missed that
discussion, and im to lazy to search trough the archives!

>But seriously, it all depends on where you make your imaginary
>"cuts" in the schemos to separate function blocks:

>All VCOs have a CCO core, of course.

Yes!

>Most VCOs drive that core from an exponential current source,
>so you have an inner block - linear CCO - and an outer block
<- expo VCO.

Yes and no, yes to inner block, no to outer block, its
confusing saying "expo VCO", say instead, "expo converter"!
Now say, im right and you are wrong!:-)

>You could also define a different block, with the cut
>at the expo converter's summing node, then you'd have an intermediate 
>exponential CCO - and the resistor to
<that summing node makes it a exop VCO. It's all just
<a matter of definition. But one thing you cannot
<do: find a cut where a partial block acts as a
>linear VCO (unless you're using the reference current
>input - which is exactly what the MS-20 does.)

But that is what i find confusing, why would the MS20 be
a more exponetial VCO then linear! To me its a linear CCO
with a exponetial driver who hapends to be driven exponetial.
The point is, for me they all are linear CCO's they just get
driven differently, that then eventually makes them "behave"
in a linear or exponetial manner. This might sound silly to
be reasoning like this but im slightely backwards!!
Thats why i do such wirdo stuff like 4046 VCO! :-)

But "what" actually are the benefits by doing what they do
in the MS20? Getting rid of a expo voltage scaling term?
Or the first order temp term?

>The Polysix and Trident oscillators are performing an exponential
>conversion that includes an expo *voltage* node. Then this expo
<*voltage* is converted to an expo current with a *linear* voltage
<controlled current source. Which then feeds the CCO core. The big
<difference to an ordinary expo VCO is that you *can* make a cut
<here such that an inner block is a linear VCO. Sure, the innermost
<block will always be a CCO, but next larger block (CCO + VC current
<source) *is* a linear VCO. So it does exist (;->) - and if you have
<offset voltage at that point you must compensate it (what Korg
<called "Low trim") - and that's also a point where you can add an
<offset voltage on purpose, which results in that infamous "linear
<detuning".

Yes that striked me to, by removing all the low trim pots and insert
a buffer who then drives all "first" 8 CCO,s one could do lin detune
by adjusting this - offset against the second row of CCOs.

But what kind of problems would a current regulator might have?
In the textbook it dosnt talk much about many, it seams to work
nice and dandy?

Korg did a neat solution to the requirement of negative voltage input
to the current regulators,by shifting the entire expo cirquit by -10 volt.
Another neat thing that came along with that is they didnt need level
translators from the CPU to the 4051 muxes.
Yet another clever design!

I did a redesign of the current regulators by using a -5volt
reference voltage the CCO sweps from 0 to 20kHz with 3Hz from
0 to-5V. If you folks would use this for LFO purpose wich i
recomend use a 10k Log poti for the freq controll.

A interesting thing with multiplexed expo converters is
that one can use sinking or sourching current regulators,
or why not driving a VCO directly, like multi vibrators!

Not to mention the combined S/H buffer and current regulator!!

>All right now - I almost went on writing "but Korg has chosen to
>do it at a different point, changing the trigger threshold of the
>CCO core." But you have tricked me - it's true they do it that way,
<and it's also beyond doubt that this results in "ordinary" detuning
<(just as in every expo VCO), and *not* in linear detuning. (At least
<not in what I called linear detuning when I coined that word.
<Ok, I see the word was misleading from the start.)

>Why is this ? You're changing the thresholds to a certain amount,
>and depending on the VCO frequency it will take a different time
>for the ramp to pass that additional voltage. Detuning is related
>to the current frequency, and not constant.

>So - nice idea, and I really believed it for a moment ! But in reality
<the Trident's detuning is quite ordinary, and when there is an additional
<linear detuning effect (and from the sound, I have no doubt about it),
<then it originates in less than accurate "low trim" - i.e. some remaining
<offset voltage in the linear VCO block. And in the fact that there is
<something like that at all - it can not happen in VCOs that consist
<of an exponential current source + linear CCO without anything
<in between.

Trick and treat! Yep, your right Juergen (looking into the shemos again), 
;->
How sad, i thought i had found somthing interesting !!! Booooooo to me!!

Hmmmm, it seams that i was "to fast" on the Trident!!! :-)
But, when i first started to reflect over the Trident way
of detune i then went to your "linear email" and what you
said was "by introducing a constant current", i think i didn't
read your email close enough. Now as it seams the "only" way
to get true lin detune in a linear CCO core is to add or
subtract current into the core itself or offset the sourced or
sinked current"  to a point in the CCO cirquit where the point
are not frequency dependent.

But then we look at the Bit 99 who offsets the second oscillators
frequency by a voltage who then are converted to a current by a
linear current regulator. So this is by offseting the voltage rater
then the current directly, but the principle are the same. Its just
converted to linear current later. So its constant current offseting
by a voltage, or voltage controlled linear detune in a linear VCO.
(if i may talk in JH language! ;->). Now this is just another example
how confusing it all is when people talks about lin or expo VCO's!!
Hehe!

>Which does not mean that a synth that would actually use lin detune
>*only* would not make sense. I made some experiments with the JH-4
<meanwhile, turning the ordinary detune as close to zero as possible,
<and switching in Lin Detune does then result in a fixed beat rate over
<the whole keyboard range. Is this desirable ? As an option, yes. But
<it's quite a restriction, too. A bit like PWM which is often used to
<sort of emulate a second VCO on single oscillator synths. IMO, for
<most applications a mix of lin and expo detune is desirable. Just
<like adding a bit (a quarter octave per volt) LFO tracking to PWM
<maybe. Modulation rate increasing with higher notes, but not doubling
<every octave. That's what I built this lin detune circuit for in the
>JH-4, and that's what (involuntarily ? or rather on purpose - who knows?)
>happens in early Moog Modular VCOs, the Taurus, the Trident, the
>CS-80: expo detune dominating, but a tiny bit of lin detuning (due to
<inaccuracies) added to add charm in the lower range.

Absolutely.

Perhapps, the Crumar bit 99 is the only machine with "on purpose
linear detune" ?? A aproach to get a interesting  string sound
is when one LFO modulates the reference frequency of one of the
oscillators. The Bit had two LFO's so interesting chorus string
sound can be achived by using separate LFOs for each OSC.

And i just "love" the sound of that machine!!

PS:The MS50 seams dated to 78-11-09.

Reg
BJ
-----------------------------------------------------------------

The End,,,, biiiip!!!!
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

Share information about yourself, create your own public profile at 
http://profiles.msn.com.