Super Klee Sequencer

[Scott Stites]

[Scott Stites]

Hmmm....maybe keep 1/6V - that's too useful looking. Dang - decisions, decisions. If anything, two muxes could be used, and one could just have some unused inputs.

The Decoder Incident is going to be anticlimactic after all of this.


G'Night.
Scott

[piedwagtail]

Scott,
I thought you had a CV quantizer....

[Scott Stites]

Hi Robert,

No hidden tuning monster here - just looking for a smaller mousetrap, but if it's too small to catch the mouse, I'll use the original, which is already on schemo

Cheerio,
Scott

[Uncle Krunkus]

1/6th is just a multiple of 1/12.
So if it were me I'd keep 1/8.

[Scott Stites]

Though an on-board quantizer could be added. External would work as well.

Cheers,
Scott

[Blue Hell]

Added a G2 demo friendly experimental patch based upon all the above, very much stripped down. See http://electro-music.com/forum/post-90463.html , no busses etc., just a 16 pot 16 stage Kle with auto feedback/random selection.

I included a little frog licking demo of it as well (with some settings that I lost later), as this seems to be a tradition with this kind of sequencer

[v-un-v]

Sounds like the bassline from 'Popcorn'- on ACID!

[Scott Stites]

Excellent Frog Lick! Thanks for that!

How many pots did you have going with that?

Last night, on the drive home, I recalled your mention of inversion of the input of the shift register. The thought occurred to me that if one strategically placed an XOR on the input of each shift register IC, a high applied to the opposite pin would invert - easily done by adding (even more) switches. This could be used to invert the action of the gate busses, too. A later model Klee, I suppose.

I've finished a rough draft of the decoder with the changes discussed. For now, I just left it at 8 voltage selections, but a CD4067 could increase that number.

The voltages I chose out of the Krunkuvolts were:

1/12 - because this is a half step increment. Mixing together would cover all possible notes. High bit density sequences would work well with it. Max note range with all bits high would be 8 whole steps.

1/6 - because this is a whole step increment. Mixing together would create whole step only sequences. Max note range with all bits high would be 16 whole steps - around 2.5 octaves, all cranked.

1/4 - because this is a half step and a whole step - would be like 1/12, but with an expanded frequency range. A bit over 3 Octave max range.

1/2V - because it's a half octave. 8 Octaves with all bits high and all pots cranked.

1V - whole octave range - would peak out at around 15 octaves with everything cranked.

2V - two octaves range per pot.

4V - four octaves range per pot.

8V - eight octaves range per pot.

Now, the idea is, on the small increments, if one wanted to, one could simply crank selected pots to max and get those increments. Of course the range on all of the the above is 0V - RangeV. And, of course, the ranges also serve to make tuning easier. The higher ranges will be more useful, generally, for 'normal' sequencing, or low bit density Klee sequences.

Problem I'm working out with the draft now is that I'm wondering if a buffer should go before each divider. Adds a few quad op amps, but looking at it, without buffers, one op amp output has 8 resistors to ground in parallel with the output - might be too much juice to expect out of an op amp. A resistor string could be made, but that would pretty much kill any idea of trimming for reasonable accuracy. Right now each divider is a resistor and a trim pot. Easy enough to trim - flip one bit on, crank its pot and tune the trimmer for the expected max voltage at that range. Click to the next range, and repeat, etc. The small increment ranges could be accurately tuned using a V/Oct VCO. Tune the first range for 1/2 step up, the next range for 1 step up, etc. This would eliminate the need for a multi-digit DMM.

Cheerio,
Scott

[mosc]

All this talk about shift registers inspired me to patch up a little patch. You can find it here: http://electro-music.com/forum/topic-13453.html

I took the SR outputs and put them into a ADC and then used that to control a 16 step Control Sequencer. This gives you the trigger bus and if you want, you can use the Control Sequencer to set the pitches associated with each value.

It should work on the G2 demo program for those that don't have a G2.

Shift Registers are more fun that a barrel of monkies. (Stupid expression, but I can't think of anything better at this time.)

[zipzap]

I´m totaly overvelmed by this post! A few days ago i finally had the time to read through all of this, understanding 1/3, ordering the shift registers.
Now some days have passed and this post has grown, so now i understand 1/5.
Anyway, Scott, this is so cool! The original first klee is really great, especially because it only needs 4 pots to play till infinity.
When the parts arrive and i learn from the early klee stages, maybe someday i´ll be able to follow up to this, hopefully.
This is like no other sequencer i know of.

[Scott Stites]

Hi ZipZap,

Mosc is right - shift registers should have just been named monkey barrels when they were first invented

The four pots is the beauty of the original Klee. Of course, I've managed to really throw that out the window with this one, but the pattern programability and gate bus make it worth it as a more ambitious project.

Cheerio,
Scott

[toppobrillo]

hey this reminds me of the first idea i had about how to DAC the matrix-scanning keyboard casio conversion i worked out. my first idea was to use
2 4051 8-1 mux/demuxes. one of which was to handle 1/12v steps, and the other would do 8/12v [because the 4051 is 8-1] steps. the voltages were derived from a current source/resistor ladder, adjusted accordingly..

JOSH

[Scott Stites]

My feeble brain can't get around deriving the voltages out of a single string and still being able to cal each one exactly with this setup, so I have eight individual dividers and an Excel spread sheet to calculate the values.

The Decoder Incident post should wait til I try this stuff out, then I'll post the final schematic and documentation, then go for Model 3.

I toyed with a cursory panel layout - Model 2 is quite a handful, but can fit on a 4U tall rack panel with reasonable spacing of controls.

Only thing I've done this week is add lag. I thought about a lag selectable between expo and linear, but I'm trying to keep controls to a minimum on Model 2, so I've settled on two linear lags, one for each voltage out (there are two output voltages - makes things much more interesting). I selected linear, because it just sounded so much betten than expo.

There's also a V/Oct input for external control either by a keyboard or another sequencer, and a modulation input. I've also put in a coarse and fine offset for the output, so one can use these controls to shift/tune the composite output. That's one handy thing about Ray's sample and hold that I always liked - the ability to offset the output without having to tie up a mixer.

Cheers,
Scott

[Scott Stites]

This weekend I've decided to confirm my schematics - IE, rebuild the Model 2 completely from the schematics I've drawn. This is a good process for shaking out any bugs that may have accumulated during the initial breadboarding and first round of schematics.

It's a good thing I did - I discovered two mistakes on the Clock and Load schematic. The schematic uses the first section of U3 twice. The second mistake is C14 is tied to ground instead of +V - that one really threw me for a loop for a bit there. I'll fix the schemo and upload it as version 1.01 to episode 1. Everything else worked out fine when breadboarding straight from schematic.

Clock and Load, the Encoder, and the Gate Bus are all singing along happily on one breadboard. Tomorrow I'm going to work out the final version of the decoder, and Model 2 will be in the can.

I've decided to add output LEDs to the gate bus - it helps to get a visual on what each bus is doing. That, and the display is mesmerising. I'm thinking of adding a function to the uncommitted half of the LM358 on the encoder. If one turns that into a pulse and makes SW17 on Clock and Load SPDT, one could have a position that randomly reloaded the Klee pattern. It's just an extra cap, with no added panel overhead.

I've also been considering trying a reduced part version of the Gate Bus. I think it could be squeezed down to six ICs from nine without losing any functionality. That will come later - I want to work out the final version of the decoder tomorrow.

Cheerio,
Scott

[Uncle Krunkus]

More fun than a pant full of ferrets!

[Scott Stites]

Progress and a weird bump in the road this weekend:

Had to dissassemble the Klee breadboard to make room for the decoder (the new Clock And Load, Encoder, and Gate bus were put on my other large breadboard). Took forever to remove the cluster of alligator leads and wires that had grown up around it just to remove it from the archipelago.

After clearing the breadboard off, and just after I laid the row of 16 trimpots that will simulate panel mounted pots, the thought occurred to me that scaling the output with an attenuator will not work as intended. The reason is this - if the output is scaled so that, when a pot is at full range, it is putting out the .083V of a half step, the original voltage will still be the 14.5V or so the shift register is putting out. That would work OK for a 'normal' sequencer, but not the Klee. Once another bit becomes active, its voltage is intended to mix with the voltage of any other active bit. If two pots are railed, that would add up to a required 29V to scale down to the .166V desired at the output. Of course, this is never going to happen, because the mixing op amp is going to rail. You'll never get greater than .083V out of the Klee sequence, no matter how many bits are active.

We can still get the voltage range selections, but the voltage is going to have to be scaled before mixing. So, back to plan A - each bit of the shift register will be dedicated to the control input of a CD4066 cell. A reference voltage will be scaled and fed to the input of the CD4066, and the output of the 4066 will be mixed with the output of all the other 4066 cells. That will require 4 CD4066s in all. I've got that section breadboarded and 16 pots. Things came up and I wasn't able to get to the voltage scaler.

Not to say the breadboarding went totally smoothly, there was the one glitch with C14 that still has me scratching my head. Perhaps someone can shed some light:

First of all, the original schematic has C14 tied to ground, which actually is the normal method of debouncing a switch. This switch is used to step the sequence forward one step each time its pressed. In this case, I could press the switch once, it would advance, twice it would advance, and around the third time or so, the entire flipping circuit would go absolutely nuts - 15V would suck down very, very low - mebbe 2V or so. To make a very long a painful story short, I finally had to disconnect both CD4034s to make things not go terribly wrong. I noticed that the second CD4034 would get very hot, not a good sign. Anyway, amongst my ministrations, I tied the capacitor high, and all troubles went away, and, oddly enough, the sequencer worked just fine stepping manually that way.

It wasn't until later, when re-doing the schematic, it just didn't look right - I don't think the 'modified' configuration is going to debounce at all.

Now, once I had everything together and working, I noticed again the second CD4034 still wasn't exactly right - its first bit on parallel load was stuck permanently high. It would shift just fine, but that first bit would never go low when it was told to go low on a parallel load. I'd never noticed before, because my initial test run had that particular bit tied high. I swapped it out, and the second CD4034 has worked just fine ever since. Now, (A) the original CD4034 was damaged when it got all hot and toasty or (B) it was defective from the get-go when I first put it in circuit (it was the first time I'd used it - the other two CD4034s were on the original breadboard).

I'm certain I had that capacitor tied low on the original - of course, I yanked that cap out of the original breadboard just to be sure I had a good cap, so that wiped out all proof, but I'm certain it was. So, I'm thinking bad CD4034. But what in the hell would cause things to pull 15V low when the event was so far upstream in the circuit? Anyone got a clue?

Hmmm....one thing I did have different, for testing, was that I had all but one parallel input bit tied to ground and the first bit tied high straight to +15 - maybe the CD4034 doesn't like that kind of twitching around when a parallel input is tied directly to the positive rail. That's not the Klee wiring, BTW, it was just for expedient testing - Klee uses pull-up resistors there.

I'll have to go back and put that cap in the initial position and see what happens now the CD4034 has been replaced. I'm holding off on rev 1.01 of Clock and Load til I try. That's a hard thing to bring myself to do, because everything is working perfectly now.....

Cheers,
Scott

[piedwagtail]

Tricky little er....swines.
I had a problem with ultra-hot HEF40106BPs as lfos.They are alright down at 5V supply with a dedicated 7805.
Robert

[Blue Hell]

[Scott Stites]

[Scott Stites]

Gak! Just listened to it. What a lousy sample for figuring out what the gate bus is doing. What was I thinking? I'll upload something a bit more expository tonight. Something with sine waves.

Today I hope to upload the corrected version of Clock and Load (just putting the right OR gate section in, a 'clerical' issue). I've got the schematic laid out for the Decoder, but I don't have any values yet for the voltage level dividers on it. A 'lost' episode will have to be played, Episode 5 - for the actual CV mixer section itself. Ran out of room on the Decoder schematic.

Cheers,
Scott

[Scott Stites]

Well no sine waves - had to kill Nauhgahide Thong to upload this sample (hitting the max on uploaded files). This one is the same patch, though a different pattern and pot settings. Not best content-wise, but it's a bit more expository of what the gate bus does.

Again, two VCO's, 2040 and late MS-20 filters, two EG's controlled by the gate bus, processing the Klee output, no overdubs, etc.

I use the gate and trigger outputs of different busses to drive the EG. One EG is triggered by Bus 3 and gated by Bus 1, while the other Eg is triggered by the master trigger out and gated by bus 2. I add and remove the signals as the sample progresses, which lends variation to the one sequence that's constantly cycling. I also am controlling one VCO with the keyboard in addition to the Klee, and that VCO is also intermittently processed by the lag circuit of the Klee. The final Klee page will have external modulation and an extra CV input (like the ARP sequencer had) which is really handy to have.

Sample also promotes a new Krunite lingerie item as well....

http://mypeoplepc.com/members/scottnoanh/sitebuildercontent/sitebuilderfiles/toadskin_teddy.mp3

Various faces of Klee sequence illuminated by gate bus outputs, best listened to whilst wearing hallucinogenic nightwear.

[Uncle Krunkus]

Very nice!
Which of these schems is the most "carved in stone", just in case I feel like starting the stripboard version in between building the Sorcerer?

[vtl5c3]

Nice track Scott. I can almost visualize a three eyed stripper dancing to it.

[Scott Stites]