Saturday, June 25, 2011

Buchla 291 Dual Bandpass Filter Module

Thanks to Scott Stites, Luka, Mark Verbos, and others involved in bringing this design to DIY and of course to Don Buchla for creating it.

The Verbos design:

http://www.simple-answer.com/291.pdf

The electro-music.com thread presents a few variations on the Verbos design:

http://electro-music.com/forum/viewtopic.php?t=14380

And Scott goes even further, discussing voltage-controlled resonance, etc.:

http://mypeoplepc.com/members/scottnoanh/slsdiy/id11.html

Marjan Urekar posts a clarified version of the original schematic with interesting notes:

http://urekarm.tripod.com/synth/download.html

Everyone gets rid of the selected 2N3958 dual FET, which seems to go for about $7.00 each on eBay as of this writing. I can't say if that effects the sonics of the module or not.

The original uses the 2N3566 transistor, kin to the 2N3565 found across the years in many places in Buchla designs. The 2N version is long out-of-production, and unless they get more, Mouser have just obsoleted the PN version of the 3566. I got a few...and it turns out that Mouser still have a lead-free version for 0.64 each:

http://www.mouser.com/Search/Refine.aspx?Keyword=pn3566

One builder in the EM thread states that he first used the 2n390X and later switched to the 3566, finding that the 3566 is a bit smoother to his ears. If so, it would be good to have modules using each type of transistor!

So the overall news. I've built Luka's version, tested and enjoyed it, then gone ahead and began replacing things to more closely approach the original component list:

-Carbon Comp resistors (not all, don't have -every- type in stock
-PN3566 transistors
-Silver Mica capacitors for the .022 pair in the original schematic

I haven't compared this to one with all modern components but am seemingly aware of a slight change in the overall character; can't say if it's for the better or worse. Will have to do a side-by-side.

For those who have read "Blink" and validated its concepts, the "feel" of the 291 did indeed change with the substitution of carbon composition resistors, toward something less clearly defined. Will be interesting to see if this is the case sonically.

Again, why pursue slight changes away from modern standards? Because I've owned the original and like it just like that, and want to re-place exactly that.

What is it like?

This filter, like many Buchla designs, mis-behaves if you're expecting utter predictability. This is "character". Using the original BPF mode, you will find that the amplitude in the bass region is something like 12dB higher than the upper octaves. A sweep down into the bass will WHUMPH you up if you're not expecting it. And as you'd expect with vactrols, it's plucky and "acoustic". Although it doesn't close completely (30Hz being listed on the front panel as the minimum setting and the 12dB response means some pass-through even then), in HPF mode it will indeed go up and out to silence (the clever engineers at Korg in the early workstations with multimode digital filters simulated an HPF fully opening by allowing the Fc to raise to the point of instability/nyquist and then lowering the volume after that; sounds like an HPF opening all the way. Kudos!).

The BPF mode makes detuned-sawtooth synth strings very simple. The wide bandwidth cuts the bass but doesn't eliminate what you need in the timbre.

A very wide-ranging, great-sounding filter. You should have a pair or more in your setup.

Changes to the modified design

The modified design offers outputs for lowpass and highpass responses, and the luka layout adds voltage processing from a CGS design. It takes up nearly the same amount of space as the Verbos dual unmodified design, which is the size of an original Buchla PCB. I'm not a fan of stacking boards if not needed...

Tuesday, June 21, 2011

Buchla 106 Mixer

I've had a gent CAD up PCB artwork for a modified version of the late 1960s model 106 mixer. Again, I await Don Buchla's permission to publicly share this new PCB artwork; I will not do so without it. Here's hoping that he will; I would enjoy publicly sharing these builds with everyone interested..

---

Edit: I've recently posted the new schematic and new artwork here:

http://www.muffwiggler.com/forum/viewtopic.php?t=65156&start=0&postdays=0&postorder=asc&highlight=106

---

I have schemos of the CBS and pre-CBS 106 mixers. Off the top of my head, the only difference is a typo in the CBS version, C8, which repeats throughout the circuit (C4 and C11), and is correct in the other schematic, and that the earlier version states 47pF caps for C2 and C6versus the later CBS version with 5pF.

http://tinyurl.com/3lksf82

Due to the roll-off in the 10uF caps throughout the design, which I'm told is at about 70Hz per cap, I've opted for 100uF throughout, which should be ten times lower.

Each section of the mixer inverts, and I wanted each section (1-3, 4-6) to have both inverting and non-inverting outputs, for various reasons (inverting AC modulation/waveforms, resonant feedback loops, etc.). I have confirmed functionality for the regular sections and that inversion works and now have to determine the optimal gain structure(s).

As stated in the schematic, I used HFE matched PN3565 (cheap at Mouser.com). I was hoping to use THAT transistor arrays but was dissuaded by the gent doing the layout owing to voltage differences. I do note that the THAT is used by Dustin Stroh (J3RK on mw.com and e-m.com) in his Verbos/258J build, for the last two 2N3565 which IIRC are involved with the square or saw shaping, and it works just fine.

I do not know if there would be any sonic differences in using other transistors such as 390X and BC55X types.

I also note that it distorts with even a single oscillator input (my 258B redo seems to have 1.7VRMS for that particular build, which is somehow above Buchla's nominal rating of 1.1). A CD player also distorts the circuit with the gain wide open. It =might= be a useful clipping in certain instances, but overall, is not immediately "musical" as some fuzz and distortion pedals at minimum settings.

I've had the 68K input resistors placed on-board to avoid lots of front panel wiring. It all fits on a single-sided 4"x6" PCB, which you'd expect from such a sparse circuit.

Why not go with a simple opamp circuit, which could also be DC-coupled for mixing control voltages? I wanted to see if the vintage, discrete circuit had any particular character (or lack thereof). It's too early to tell but yes, I'll later add a DC mixer circuit to this circus...

And I am indeed using AB resistors in all of the stated places. No photos as I don't own a camera or equipped cell phone.

Tuesday, June 14, 2011

Carbon Composition Resistors in Buchla Gear

The 258C seems to have random population of CCs, but the 258B schemo specifically indicates certain values to be carbon comp:

1, 2.2, 4.7 multipliers (10R, 1K, 10K, 100K, 10M, 2.2R, 2.2K, 470K, 4.7M).

In the last post I note that the 158A indicates that some resistors should be 10%.

"On the CBS schemo, note that some resistors are 1%, and some are specifically listed at 10%. These are at a glance, 1.0, 2.2, 4.7 multipliers, and include the 330K as well (which also appear in the sine shaper). Unremarked resistors are stated as being 5%. I note that the 258B schematic lists all 1.0, 2.2, 4.7 multipliers as being specifically carbon composition types, all others tin oxide. I wonder if you looked at a 158B/CBS/Tape Music Center unit if all of these would be Allen Bradley or other carbon comp."

I also note that the CBS 106 mixer schematic lists many resistors as 5%, and the rest as being 10%. Again, these are the 1, 2.2, 3.3, 4.7 multipliers (1K, 10K, 220R, 2.2K, 330R).

I've asked a 158 owner for a pic to see if they are there...

If you are on prodigy-pro.com/diy or other higher-end gear pages, you will note that carbon composition resistors are spoken of highly. I note that on audiophile DIY sites that it goes further, with tantalum resistors being the item of choice in certain critical places in the signal path, especially in circuits with gain as it is said that tantalums have no thermal noise, as do metal film resistors. Then again, tantalum resistors seem to start at about $3.85 each, and go up to around $10.00 each.

I note that since these are vintage components, they can suffer from slight oxidizing on the leads. A slight rub with loose steel wool right up at the resistor body will clean these up quickly.

Saturday, June 11, 2011

Buchla 158A, 158B, 158 CBS versions

A thread in the muffwiggler.com DIY forum brings up interesting questions:

The CBS version up at Magnus' Buchla Page is extremely similar to the B version.

A version:

-All discrete, no opamps
-No uA726
-Transistors: MD708B (dual), 2N4916, 2N3565, U147, E101 (E101 is selected for =approximately= .5 Idss).

http://www.semi-tech-inc.com/viewsemi.asp?STI_Type=MD708B&tbl=table9

The E101 also appears if I remember correctly in the Minimoog oscillators, and the U147 also appears in the earliest versions of the Moog 901B oscillator.

B version:

-Opamps: LM302, LM301A, uA726
-Transistors: 2N4916, 2N3565, 2N4339 (Idss between .7 and 1.2), 2N4339 (Idss between .5 and .8).
-D1-D3 are listed as FD111 (???)

On the CBS schemo, note that some resistors are 1%, and some are specifically listed at 10%. These are at a glance, 1.0, 2.2, 4.7 multipliers, and include the 330K as well (which also appear in the sine shaper). Unremarked resistors are stated as being 5%. I note that the 258B schematic lists all 1.0, 2.2, 4.7 multipliers as being specifically carbon composition types, all others tin oxide. I wonder if you looked at a 158B/CBS/Tape Music Center unit if all of these would be Allen Bradley or other carbon comp.

As mentioned elsewhere, and thanks again to J3RK, the J201 transistor, easily available, has an Idss range which falls easily to .5 or so. I'd love to learn about the others and what would sufficiently replace them. The 2N4339 doesn't seem to dip down into that range very often; my batch go down to about .66 as a minimum. Useful for the 158B..

I note on the 158A part layout document (5Apr68) that there are a few bits not mentioned in the schematic:

-Note- "Plug in" version.
-R8 Sine: 1.5K; Harmonic: 680R.
-Add R34, (other notes, too distorted to read)
-No C7

There is a fat pen mark at the upper right, stating 158A-L.

I don't have the schemo in front of me so I have no idea of what "harmonic" means.

Wednesday, June 1, 2011

Oscillator Drift Iteration

Ye Olde Oscillatore Drifte question...

The Buchla 258B, with the 2N3802 dual transistor in a single package and the uA726 heated transistor pair (again in the same package/substrate) provide a very decent amount of pitch stability without going into the land of totally uninteresting. I wanted to see what component variation would cause/effect.

This especially centers around the removal of any dual transistors in a single package. Using modern transistors, matching isn't terribly hard for the 1% and better range, but not having them in a single package/substrate provides for less overall accuracy. Having them physically contacting each other with heat grease can improve things but in some designs such as the Moog 901a/b, these transistors are not placed close enough to each other to allow for physical contact.

Individual transistors with no physical contact per pair seems very important.

-The 726 and 3802 I have replaced with selected BC550C and 560C and of course, find myself in maximum driftland. Adding a TEMPCO noticeably reduces this but does not eliminate it. Might be nice to have a single module set up this way. Didn't even bother to heat grease them together to see if slightly less drift could occur.

-Tried a TEMPCO plus uA726 plus 2N3802 and it powers up and STAYS there. "No, too much!" -The Who, "Magic Bus", "Live at Leeds"

-I note that the J3RK/Verbos 258 design using the THAT340 is stable enough without a TEMPCO. I hope to have a future iteration of the 258B design using a THAT340 to replace both the uA726 and 2N3802. I would have pads and components added so that you can use the vintage parts instead, if desired.

The J3RK design uses the second half of the 340 IIRC to replace the 2N3565s. A tech tells me that the THAT has a quarter or so of what is needed to correctly operate in place of a 3565 pair...could be interesting to compare how 3565 and THAT versions sound/look on a DAW.

-There is indeed a certain solid "feel" to the uA726; it grabs the pitch nicely but not absolutely.