alternatives to hartley or colpitt oscillators?

[parmalee]

First, this is not a stompbox query, but it's certainly not unrelated.

I'm recreating the bass section from the Farfisa Compact Organ for use in my mini-modded-pedal-harmonium setup,with sundry effects (don't ask), and the topology of the master oscillators seem to be of the Hartley or Colpitt variety, as far as I can tell.  Fortunately, as  I'm only doing an octave's worth of notes, I needn't  worry about multiple dividers for each oscillator; rather, I just have to scale the master oscillators down to the 32 - 64 hz range.  Interestingly, the bass pedal (or  bass section—bottom octave of keyboard) tones are just the bottom octave tones of the 16' tab processed through their own unique filter and preamp section--and it is this latter stage which provides the "magic" that is the Farfisa Bass Sound.

Anyways, I'm wondering if there is another fairly straightforward transistor oscillator topology which will allow me to use basic axial lead inductors and a tuning capacitor perhaps (or even no inductors at all--like a UJT oscillator?), rather than the cumbersome—and rather scarce these days, except when salvaged from combo organs—"tuning coil" setup as shown in the schematics below.  Any thoughts?  (The master oscillator surrounds that transistor all the way at the bottom of the pic--the rest are the dividers/flip flops)

[PRR]

CMOS Schmitt-trigger R-C oscillator will do the voltage and frequency.

Problem is... most of the sweet simple cute oscillators will drift. Badly by musical standards.

L-C oscillators can be much less drifty. A small change of frequency is a large change of phase so they can't drift far. Drift is a problem in radio so designers knew a lot about the drift of coils, and had selection of caps to compensate the coil and stray drift.

However down in the low audio band the coil(s) will be HUGE. Values of several Henries. With practical copper and iron, either you have so much stray resistance that the coil is not very coil-like, or you add iron which makes the actual value even more uncertain.

The cost of the huge coil will surely exceed the cost of a dozen divider stages.

[PRR]

BTW-- I don't think that is either a Hartley or Colpitt. And I do not think it matters what "name" you put on it. Back when oscillators were novel, writing-up every different detail and naming it after a person was a big deal. And things were simpler then because a filament tube only had two non-grounded terminals. With cathodes and especially with transistors, you can connect all different ways and nearly all of them will oscillate under some set of values.

I'd think of it as "tuned plate with grid tickler", and maybe there is a name for that, but knowing the name is not a lot of help.

Ponder http://www.electronics-tutorials.ws/category/oscillator

[PRR]

Your plan, in essence, is found 2/3rd of the way down this page:
http://www.electronics-tutorials.ws/oscillator/oscillators.html

Their sketch omits necessary biasing, and some parts that improve stability by not allowing it to implode when it clips. But of course ALL oscillators can be described by the math above that example.

[parmalee]

Posted by: PRR

Your plan, in essence, is found 2/3rd of the way down this page:
http://www.electronics-tutorials.ws/oscillator/oscillators.html

Their sketch omits necessary biasing, and some parts that improve stability by not allowing it to implode when it clips. But of course ALL oscillators can be described by the math above that example.

Thank you very much for this link.  Oddly, I've been perusing that site for a while now but did not manage to stumble upon that page.  Much more comprehensive.

I had kind of wanted to do this discrete--not with germaniums, mind; silicons (as used in the Farfisa FAST series: FAST=Farfisa All Silicon Transistors) are fine.  But it appears as though doing such will be much more involved.

Funnily, the master oscillator in the Compact and Professional series produces something more akin to a triangle wave--but then it goes through all the flip flops for the lower octaves, rendering all square waves:  So what's the point?

When I mentioned the Hartley and Colpitt I was actually thinking of the Professional's master oscillator (too much work to reproduce here--it's in a big PDF).  By that stage--'68 or '69--Farfisa had shifted from all PNPs to all NPNs--any idea as to why that might be the case?

Do you think that the sound will be substantively different if I go the CMOS Schmitt trigger approach?  I can still reproduce the subsequent filter-and-preamp stage as per original, and I do believe that to be the source of the "magic."  Otherwise, perhaps I can source a dozen or so older tuning coils from a Vox Jaguar or Farfisa FAST or VIP organ relatively cheaply--they are also far less clunky to boot!

[anotherjim]

I'd go with a CMOS schmitt. Well regulated supply voltage essential and a convenient tuning control - perhaps you can use a type of trimmer fitted with a small knob . It will drift, but organs existed with eccles-jordan astable oscillators which were no better.

Also, you could take the master oscillator route. There is a way to replicate the old top octave chips with cmos counters I've seen somewhere.

Would be interested in seeing the Farfisa bass filter. Given the liking for octave down effects using flip-flop dividers around here, it could prove quite useful in a stompbox 

[amptramp]

If you have the space, you could build a transistor equivalent of the original Hammond Solovox:

http://antiqueradio.org/art/Solovox-technical.pdf

I have the later version, the Model L.  An L-C oscillator has a frequency based on 1/(L*C)^0.5 whereas an R-C oscillator has a frequency based on 1/(R*C).  This means that a variation in L or C only has half od the frequency drift effect of a variation in R or C of and R-C oscillator.  A Vackar oscillator is designed to have large capacitances in parallel with the active device, meaning that active devices with capacitances that are all over the place will still work.  The Vackar oscillator is defined here:

http://en.wikipedia.org/wiki/Vack%C3%A1%C5%99_oscillator

Other examples are here:

http://www.qsl.net/va3diw/vackar.html

It has been used for ham radio oscillators for some time as variations in active device capacitance have little effect on frequency since device capacitance is swamped by circuit capacitance.

A unijunction oscillator would work like a 555 oscillator.  There are other topologies like the comparator oscillator that would be similar and would have the advantage of having both triangle and square wave outputs.

[R.G.]

The bottom line on this this much as it's been stated: R-C circuits do not have enough stability to be usable for musical sources unless you're willing to tune them to pitch before each use, much like guitarists are always reaching for the tuners. Musical notes have very tough requirements for stability, both in the short run and long run, and over temperature ranges.

You could use high-stability resistors and caps if you do something like incorporating a $5 chromatic tuner to tune it up before playing. That may be the best way.

I went through a bunch of this back when I was fascinated by the great highland bagpipe scale a ways back. I finally decided that for my uses, a crystal controlled $0.75 microcontroller that could generate a stable note at any musical frequency was just fine, and simpler than the tweaking and parts sourcing issues. But that doesn't help you much if you want it to be all discrete.

[Mac Walker]

Here's a $5 top octave generator (with a little bit of coding of course)

http://hackaday.com/2014/06/16/cypress-launches-5-arm-dev-board/

Best I can tell you don't need to spend $8 billion on a special cable or development/interface tools....

[R.G.]

That development board would probably do it.

I got frustrated with the top-octave approach to making musical notes - too hard to get all those notes with non-integer and non reduceable dividers inside one uC. I realized that if I sliced the octave banks vertically instead of horizontally, it gets easy.

One PIC has a hard time making the top musical octave. It is trivially easy to get ONE note of the top octave, and all 8 lower octaves in one PIC, though. So instead of one top octave generator IC and 12 divider chains (like a CD4024 per), you have one PIC generate all of the C; another all the C#/Db; another all the D, etc. It actually uses one less chip as the top octave generator is no longer needed. The cost of a baby PIC that can do this is under $1.00 in ones, way under $1.00 each in quantity. So a generator for all the musical notes can be had for $12 and one crystal, about $0.50. Of course, it still costs you that for only the bass notes, but what you gonna do? 

[amptramp]

I once rebuilt a Minshall Model E organ with an octave generator consisting of 12 Fairchild µA2240 devices:

http://www.datasheet-pdf.com/datasheet/TexasInstruments/524769/UA2240C.pdf.html

This is a 555 timer configured as an astable followed by a chain of eight dividers.  It worked, but only with Korean IC's , not the Malaysian ones.  The woodwind voicing was excellent (since a clarinet sounds like a square wave generator anyway) but the brass voicing was missing something (since it is triangular wave).  You could design a driver where the top octave had one oscillator set to 440 Hz (key of A) and was divided by 185/196 divider stages driving PLL generators like a tone wheel generator on a Hammond organ.  Tedious in random logic but easier with a PIC (or several) or some more advanced processors that could do this on a chip.

Since a number of oscillators operating off the same power supply may tend to injection lock at each other's frequency, there is a solution for this: the phase unlock loop.  This guarantees a large bump to the VCO control voltage to keep oscillators from locking into each other's frequencies.   

[parmalee]

Well, I might have played up the "filter and amplifier section" a bit, but it can be found on the last page of this PDF--marked "1/2 PA 10" (the other half of the board is the vibrato section):

http://soundandcircuit.webs.com/Farfisa_Mini_Compact_Service_Manual%20%282%29.pdf

A couple of RC low pass filters and a couple of common emitter amplifiers (if they were NPNs, what are the PNP equivalents called?) seemingly provides the "magic."  Well, that and a suitably powerful guitar or bass amp and cabinet.  The tone color circuitry for the rest of the organ is pretty interesting though, especially Farfisa's unique "Multi Tone Booster" section:  operated by a knee lever, and employing a couple of lamps and LDRs--wah and volume pedal-style--it uses a high pass filter and sounds pretty awesome considering how minimal the circuitry is.  This can be found on page 2 or 3 of the link above, I believe.

I bread-boarded a mock-up of the filter and amp section (using NPNs though) and ran the most basic Schmitt trigger (from a 4093) through it last evening and it sounded pretty close to what I am shooting for.  I am a trifle concerned about the instability of this approach, however--perhaps I can source a cheap Top Octave Generator chip from a 70's or 80's organ model, and divide accordingly.

I do believe that part of the "magic" to which I alluded (and I'm not crazy here--many have remarked upon the exquisite beauty of Farfisa Compact bass tones) is simply the fact that, being an organ and not a mono synth,  it's polyphonic and when playing a bass line in a legato style there is a very brief moment when two notes sound simultaneously.  No matter how amazing a bass sound one can get from a mono analog synth (keys or pedal board), it's always missing a certain... something that I get from a Farfisa Compact bass section.

Posted by amptramp

You could design a driver where the top octave had one oscillator set to 440 Hz (key of A) and was divided by 185/196 divider stages driving PLL generators like a tone wheel generator on a Hammond organ.  Tedious in random logic but easier with a PIC (or several) or some more advanced processors that could do this on a chip.

This sounds interesting.  I've often contemplated electronic emulations of tone wheel generators, but a full-scale undertaking of such without PICs would be more than a little tedious.  And contrary to my current thinking, which has me aiming for fitting everything into the smallest enclosures possible.

[parmalee]

If you have the space, you could build a transistor equivalent of the original Hammond Solovox:

Just out of curiosity, roughly how much does your Solovox weigh?

I haven't heard many recordings of the Solovox, but Charles Hayward used the Maestrovox on the This Heat albums (especially the first one, (Yellow/Blue), and it can be heard on a number of their live recordings, as well.  Most of the info I've found on the Maestro is on this site: http://www.debbiecurtis.co.uk/id114.html.

[parmalee]

Anyways, after some consideration I'm still leaning towards the individually tunable notes route, as opposed to utilizing a TOG (or TOS) chip--although if a crappy 1970's Kimball or Lowrey spinet happens to fall into my lap, I won't turn it down.  It's just that I've been living "in the middle of nowhere" for the past couple of years, and I don't find organs and such on my doorstep anymore like I did when I lived in cities--now, I have to go to them!.

But the comment on the tone wheel generator approximater got me to thinking just how much I like a little bit of "inharmonicity."  I used to deliberately tune one divider board--probably the F#--off by just the tiniest bit--like a couple of cents.  It was virtually unnoticeable when playing the higher organ notes, but ever so "subtly" noticeable when playing the bass section, especially polyphonically.  My microtonal excursions never got as sophisticated as those of Terry Riley, LaMonte Young, or John Cale (or any of millions of non-Western musical performers), but nevertheless, doing such added a bit of color.  And it was psychologically unnerving to some, which made it all the more enticing.

[PRR]

> By that stage--'68 or '69--Farfisa had shifted from all PNPs to all NPNs--any idea as to why that might be the case?

1950s into the 1960s, Germanium was king (Silicon was hard to work). I don't know the exact reason, but in Ge the PNP is easier (cheaper). And when complementary circuits were rare, NPN types were rare. "All" designs started as all-PNP.

Early 1960s, Silicon was highly favored for computer and military work. Fall-out from mature Silicon production lines became the jelly-bean transistors for audio. For some other reason I do not know, the easy Silicon device is NPN.

Your organs have 144 transistors just in the part you show, and more to get the signal out. This is a l-o-n-g way from a 4-transistor radio (yet far short of a 30,000 transistor computer). So organ production started when Ge types came well below $1/each. Because of drift and shifting supplies, designs moved to Silicon in some mid/late-1960s re-design. Either way the economics strongly favor the polarity (PNP/NPN) which is cheaper at the time.

As you know, today small PNP and NPN Silicon are "the same price". The actual gut is now such a small-small part of the cost of getting one to you that the price difference is lost in factory - warehouse - wholesaler - retailer overhead. But I sure remember NPN being a dime cheaper than PNP. (A whole buck+ for power devices.)

[amptramp]

If you have the space, you could build a transistor equivalent of the original Hammond Solovox:

Just out of curiosity, roughly how much does your Solovox weigh?

I haven't heard many recordings of the Solovox, but Charles Hayward used the Maestrovox on the This Heat albums (especially the first one, (Yellow/Blue), and it can be heard on a number of their live recordings, as well.  Most of the info I've found on the Maestro is on this site: http://www.debbiecurtis.co.uk/id114.html.

Always interesting to run into something I hadn't seen before like the Maestrovox.  I weighed the Solovox.  The keyboard section weighs 15.0 pounds.  I rebuilt the tone cabinet with different transformers because the power and output transformers were missing when I got it, but my version using different transformers weighs 24.0 pounds, both weights to the nearest 0.5 pounds.

Here is a good restoration article on the Solovox:

http://antiqueradio.org/HammondSolovoxKeyboard.htm

Rummage around that site and you will see he has also restored a Gibson Clavioline and he includes the important schematics..

[anotherjim]

On "microtuning". Our pitch perception goes a bit weird at the extreme ends - dependent on volume to some extent, but isn't this behind so called "stretch" tuning on pianos? There are modern recordings using sampled bass that sound a bit off, but not apparently noticed by the producers.

Overlapping bass notes -  I used to have a Logan String Melody string synth that had a percussive bass voice. There was a glitch I never got to the bottom of where only a few adjacent notes would hang on a bit and there would be a lovely Tuba like growl when they briefly overlap.
The same thing happens in highly reverberant spaces such as with church organs.

[PRR]

> isn't this behind so called "stretch" tuning on pianos?

The prime reason is that the overtones of a string are never perfect integer intervals, because the string stiffness is significant and more significant for the shorter wavelengths.

The effect is slight on guitar. But piano uses larger gauges. The overtones are way sharp of ideal intervals. To keep from sounding too bad, the fundamentals of higher notes are "stretched" to fall nearer the overtones of lower notes.

Rhodes tine-piano has a tapered section to approximate the inharmonic overtones of a real piano.

You don't stretch an organ/synth (unless emulating a piano) because their overtones ARE exact integer intervals.

[amptramp]

There are some organs (like Thomas) and some synths that use exact harmonics but the results are not that realistic except for woodwind voices and even for them, more complication can get a more realistic sound.

If you watch music on a scope, there are patterns that appear to run backwards and forwards slowly along the waveform and this is part of the identity of the sound.  For example, an acoustic guitar, the fundamental may be at 300Hz, the next "harmonic" at 595.4 Hz, the next at 897 Hz, the next at 1191.8 Hz and the next at 1500 Hz.  The fifth harmonic is exact but the others are low due to the detuning effect of the hole (this is an acoustic guitar).  If you follow the waveforms on the guitar string, there is little coupling from either the fundamental or the fifth to the hole, but there is enough coupling to detune the other harmonics (some people call them "partials" because they reserve the term "harmonic" for exact integers.  The energy lost to the hole decreases the frequency of the other partials just like loading a tuned circuit will cause its frequency to decrease.  For a guitar, the partials are at multiples of 1.00 (fundamental), 1.985, 2.99, 3.994 and 5.00, not 1, 2, 3, 4 and 5 as would be generated by a phase-locked synthesizer.  The energy lost is proportional to the energy coupled to the hole.  If the waveform along the string has a null over the hole, little energy is lost.  If it is at a peak over the hole, a lot is lost and the partial decreases in amplitude faster and the frequency is lower.  These patterns that run along the waveform show up at the difference in frequency between an exact harmonic (which would be stationary) and the actual partial.

Bowed instruments have a different physical mechanism: the bow stretches and adds stress to the string until the string overcomes the static friction and snaps back into place, so you have a constantly changing "fundamental" and partials and the side of the string beyond the bow behaves differently from the side of the string nearest the player, but they both lose friction at the same time, so you have a nice set of differential equations that I don't want to solve.  This is best simulated by FM (frequency modulation) synthesis where a sine wave is frequency modulated at a rate slightly off an exact harmonic to generate a waveform with non-integer partials.  The Yamaha DX-7 was the first popular FM synthesizer and it was known for being able to generate realistic voicing with the limited electronics of the era.

Most instruments have an "attack" frequency that settles to the steady state after wild frequency swings.  Brass and bowed string exhibit this the most.  An accurate synthesizer would have to take this into account.