DIY Top Octave Generator

Scott Stites

Another thing that I think would sell, but it totally beyond my ability, but might be easy for other members to implement (I don't know):

A TOG (Top Octave Generator) board/IC. I would think there would be a way to do it with one of them there programmable ICs. That whole area of synthesizers has just DIED along with the disappearence of the TOG IC. It would reintroduce something to Synth DIY that is practically unobtainium these days.

It seems to me organ repair centers do sort of the same thing, but they do it in the price range of the valued limb/first born.

Nothing fancy, just something that divides down to a high octave of twelve steps. Either a programmed IC, or a little board for it. I think that'd sell once people realized what they could do with it (for one thing, use JH's Solina thing to make a nice little string machine). Paraphonic would be back, baby, and that would be a good thing (IMO).

Uncle Krunkus · Posted: Sat Oct 04, 2008 3:35 pm Post subject:

Great ideas Scott,
Although I'm very slow with actually getting around to doing anything, both of these ideas have featured in my head soup over the last year or so. Well,... in a round about kind of way.
The first one sounds like a distant cousin of the Graphic ENV generator I told you about. (For everyone else, it's a collection of vertical sliders which are ganged together to form a custom ENV of however many stages you like)
The second one is related to the discussions Loss1234 and I have had about deriving chromatic tones by multiplying up and dividing down by various integer values. It's obvious to me now, both from what Scott just said, and Ian's new wavetable VCO, that just simply starting from a VHF signal and dividing down is the way to go. Of course, I'd prefer to implement it in common CMOS chips rather than a PIC. I just like them better. They take up more space, but present a lot less hard work IMHO.
I think I'll get a spreadsheet happening today!

Scott Stites · Posted: Sat Oct 04, 2008 6:49 pm Post subject:

You're preaching to the choir about CMOS Laughing

However, this is one of those cases where you'd have one hell of a huge board just to get the frequencies you wanted (there's a lot of dividing going on, and I'm not all that convinced multiplying signals is the best way to go).

You'd want something that started an octave or two above hearing, then divide down from that. That would be plenty of CMOS to wrap your iron around if you think about it. The waves would be pulse waves, but you could think about constructing sawtooths out of those waves as they divide down. Methinks some of the TOG synths did just that, with low notes being rich in sawtooth, and becoming more pulse-like as the frequency went up (because there were fewer pulse waves available to construct the harmonics for the sawtooth).

Another way to go about it, without TOG, would be to make standard boards that divided a specific, divided down note range. JH did this with his PS3XXX clone, IIRC. That would be a good selling point in a way - you would need twelve of them to construct a paraphonic synth.

I'd kill for freely available TOGs, though.

Cheerios,
Scott

blue hell · Posted: Sat Oct 04, 2008 7:44 pm Post subject:

Scott Stites · Posted: Sat Oct 04, 2008 8:49 pm Post subject:

I was thinking about the scale that fits on the white and black keys Laughing

I do have an old copy of Polyphony that covers even temperment using CMOS following a TOG, and that portion of it isn't tiny. Multiplication with CMOS can be tricky in certain instances, but if you have something to divide down, it wouldn't be an issue.

I was hoping FPGA's had come far enough? Division beyond that could be done with standard CMOS - a lot, but not an awful lot. You'd need to divide down six to eight times per note in a scale. The right counters would help to keep the parts count low.

Uncle Krunkus · Posted: Sat Oct 04, 2008 9:50 pm Post subject:

You'd need a separate set of counters for each note yeah?
So, say you start with 14080Hz, you get top A for free.
Don't you just use a couple of 8 bit counters for each note? Set a certain number through diodes to reset. And take the signal off whatever MSBs you like the sound of yeah? 2^14 is 16384, so you only need a 14 bit counter each. That's an 8 plus a couple to make up.
Have I missed something?

Scott Stites · Posted: Sat Oct 04, 2008 10:11 pm Post subject:

Uncle Krunkus · Posted: Sat Oct 04, 2008 10:55 pm Post subject:

Here are the frequencies down from a high A: -
A - 14080.0000
G# - 13289.7503
G - 12543.8540
F# - 11839.8215
F - 11175.3034
E - 10548.0818
Eb - 9956.0635
D - 9397.2726
C# - 8869.8442
C - 8372.0181
B - 7902.1328
Bb - 7458.6202
A - 7040.0000
If, for example, we reset at 13289, that's G accurate to 5 decimal places! That's a lot closer than horseshoes. How much more accurate should it be?
Most tuners don't go to 1/100th of a cent.

BTW I think we should split this off from the EM-PCB thread. Can I do that?

Uncle Krunkus · Posted: Sat Oct 04, 2008 11:48 pm Post subject:

I just checked the numbers, and if you bite the bullet and go for 2*8bit counters per note, then your division power goes up to 65536.
So, to make the most of it, take it all up by two octaves and two semitones to the B at 63217Hz.
But,....
I just checked, and we have a 14bit counter in the CD4020! So, if we stick with the "just inside audible" top note of B at 15804Hz, we've got the top octave from just 11 chips! and a shitload of diodes!! Rolling Eyes

Come on Scott!? This could be fun! Laughing

Scott Stites · Posted: Sun Oct 05, 2008 12:16 am Post subject:

It's late here so i may not be thinking the clearest right now, but reset at 13289? Wouldn't you be dividing 14080 by 13289 if you did that?

Wouldn't you want to reset every time the 14080 frequency counted 1.0595 steps or something along those lines?

Uncle Krunkus · Posted: Sun Oct 05, 2008 2:00 am Post subject:

No, actually you're right. Embarassed

I've just completely mixed up what the hell I was trying to work out!
Sometimes my brain just doesn't work as well as it used to.
Maybe it's the drugs! Shocked

What I was thinking is completely wrong. So utterly wrong I should wipe these posts! And, worse than that, it shows that I have absolutely no idea what the problem actually is. Still don't.

Uncle Krunkus · Posted: Sun Oct 05, 2008 2:07 am Post subject:

Okay,
I think my brain fart has subsided.
Next question, what's the highest frequency you can clock a 4020 with?
I'll see if I can find out for myself, but at least you know I'm thinking about it the other way up now. Laughing

blue hell · Posted: Sun Oct 05, 2008 5:48 am Post subject:

Not chop it off to a public place then?

3.5 MHz @ 5V upto 12 MHz at 15 V.

Scott Stites · Posted: Sun Oct 05, 2008 6:58 am Post subject:

If I had a nickel for every time my brain farted, I'd own this board. Laughing

Anyway, now you're thinking. This is why I still wonder if an FPGA didn't have enough gates to crunch out the hard divisions. If not, I think you could with a fist-full of counters. It would be a board that replaced an IC, but, hell, you can't get the ICs any more, so.....

magic.PNG

Description:

Filesize:

40.14 KB

Viewed:

12721 Time(s)

Uncle Krunkus · Posted: Sun Oct 05, 2008 2:59 pm Post subject:

Uncle Krunkus · Posted: Sun Oct 05, 2008 3:35 pm Post subject:

Kassen · Posted: Sun Oct 05, 2008 6:24 pm Post subject:

Ok, I feel guilty now... I've been using 2Ghz and 64bit floats... and using it to tune to integer multiples of 200Hz... and really liking that scale. Sorry. :¬)

blue hell · Posted: Mon Oct 06, 2008 4:39 am Post subject:

Integer multiples of 200 Hz sound good because it tunes well with the mains hum I guess ... and 440 Hz is obviously a mistake ... or is 50 Hz the mistake ...

Uncle Krunkus · Posted: Mon Oct 06, 2008 5:12 am Post subject:

I reckon they should split the difference and change to a standard of 55hz which wouldn't hurt much existing equipment, and is 3 octaves below A @ 440. Then we might finally "teach the world to sing, in perfect harmony! Grow apple trees, and honey bees, and la la la la la,........" Laughing

BTW Can you still chop this off Jan? From the bit about TOGs onwards. Maybe put Mosc's on topic post back afterwards. Oh, and I'll have a coffee too while you're at it, and are there any chocolate biscuits? Very Happy

Uncle Krunkus · Posted: Tue Oct 07, 2008 3:38 am Post subject:

Scott,
what ideas have you got about the master clock?
Is there an easy way to hook up an 2, 4, or 8Mhz crystal to a chip and get a nice clean clock out? Could you tweak the frequency on a crytal based clock generator like that?
Does the first note out of a TOG have to be any one in particular?
I was thinking about the 40103. Using a programable down counter such as that would simplify the board heaps as you wouldn't need to do external testing with diodes and/or gates.
I can't see how to up the bit count to 9 or maybe 10 without using external testing, which will mean gates, and diodes, and a lot more complicated layout. Unless we can find a 10bit programmable counter, or make the 40103 act like one?
Maybe it is the perfect place to go for the simplicity of 8bit for this reason, but I was kind of thinking that if we go to the hassle of making a DIY TOG, then we should make it better than the obsolete ones. What do you think?
I'm still working on finding the "magic" 10bit dividers. Maybe they don't exist, and then going up to 10bit would be a waste of time. Confused

I'll attach the spreadsheet I've been using, maybe you'll get inspired. Very Happy

Note Ratios TOG.xls

Description:

Download (listen)

Filename:

Note Ratios TOG.xls

Filesize:

62.5 KB

Downloaded:

262 Time(s)

Rykhaard · Joined: Sep 02, 2007 Posts: 1290 Location: Canada

This might seem quite silly but a year or more ago, in my investigations towards building my own string machine, I played with generating squarewaves from the simple but useful CD40106. Smile

It were stable! I'd set one up with 2 or 3 individual oscillators at different frequencies and they held their frequencies over a few hours, with no troubles at all.

1 x CD40106 = 6 different frequencies? That'd only be 2 of them for an octave. ??? Smile

For divide down - have your entire top octave done with 2 x CD40106's and then divide by 2 for each note of each octave below / etc.

No? Smile

(What's wrong with this thought line of mine. I'm not awake enough yet, to figure any problem out. Surprised

)

EDIT: A few minutes after having typed this and I remember that out in the garage, I have the Farfisa VIP 400 that I still have to bring in to the studio - give it it's missing fuse and fuse holder and there's my 'string machine'. Doh. Surprised

richardc64 · Posted: Tue Oct 07, 2008 7:46 am Post subject:

(Where was this thread split from?)

A board of a dozen or more ICs isn't going to be much of an improvement over the original 50240 and it might be worth it to accept the exorbitant price the organ repair co's want for a replacement. That, or persuade someone with the skills to provide a CPLD or FPGA recreation.

btw, 8-bit division is definitely not good enough and 9-bits is only "better."

That's surprising and good news about the 40106 osc stability. All could be tuned simultaneously with techniques discussed in the Lunetta forum.

Rykhaard · Joined: Sep 02, 2007 Posts: 1290 Location: Canada

urbanscallywag · Joined: Nov 30, 2007 Posts: 317 Location: sometimes

Scott Stites · Posted: Tue Oct 07, 2008 10:59 am Post subject: