Favorite designs

Fri Apr 18 01:05:16 CEST 1997

     Good response. To this I would add the following:

     Synth chips are best used in either polyphonic systems or machines 
     that otherwise rely on their accuracy, repeatability, and unit-to-unit 
     matching. To use them as a parts-saving tactic in DIY machines is a 
     bit of a waste, if there is a scarcity of them. Perhaps this will 
     change when the new parts runs become available, but for now, that's 
     my feeling.

     For a monophonic knob-laiden performance-oriented lead-line solo voice 
     analog speaker-blower, I'd much rather use hand-built discrete 
     circuits. Yes they may need constant tweaking, but any good musical 
     instrument (electric or otherwise) will. For this you're better off if 
     you know the machine well, inside and out, but this is what we 
     traditionally refer to as "virtuosity" anyway, right?

     So it depends on what you're after, I guess, to determine which design 
     you should go for. Here's some views, from my current thinking (which 
     of course changes over the years, so take it all with a grain of 
     salt):

     1) VCOs - The 3340 is pretty accurate and drift-free yet somehow 
     sterile-sounding, so the common wisdom goes. Reserve these for 
     polyphonic machines, especially where auto-tune is used. In the 
     discrete world, you can get some very sexy VCOs by using a mixture of 
     some old Electronotes ideas plus a careful choice of modern components. 
     The MAT series of matched transistor pairs from good ol' Analog Devices 
     will solve all of your exponential converter worries with the exception 
     of that pesky little tempco term, and if that last little drift factor 
     bugs you then you *can* find the resistors if you're really determined. 
     Both the 3080/3140 tri/square VCO and the 3140/311/FET VCO designs are 
     as good if not better than any commercial VCO you may care to name, 
     including the Serge. To this day these parts combinations still work 
     well together.

     2) VCFs - They're all different. Revel in their diversity. The Moog 
     ladder is beautiful. A good stable multimode is always fun. The 
     cascaded 3080 4-pole actually sounds very smooth. The 2040 causes 
     extreme pleasure. etc. etc. etc.

     3) Envelopes - the 3310 is God's gift to the classic ADSR world, 
     agreed. Really that is my favorite ADSR - wide range, accurate, all 
     that. And hard as heck to find. It's actually a great challenge to 
     build a discrete ADSR that will have the same specs, even without 
     voltage control. But with a discrete you can compromise something to 
     emphaisze something else, so you gotta know what your goal is. A 
     general-purpose design will always leave something to be desired. I've 
     found that a 555 makes a good A/D flip-flop element, and combined with 
     a 4053 you can make a simple yet very snappy ADSR that will fit into 
     most solo synth applications. For real long time constants you may 
     need to go another route.

     4) VCAs - Once again I agree that the chips have nice numbers. But 
     darn it, the 3080 designs sound fine! Screw the numbers, a 3080-based 
     final VCA is quiet enough for most applications, especially DIY 
     monosynth projects. The cost/performance ratio is very, very high 
     here, and that's a big pedestal to knock over. (Maybe my 3080's are 
     top-quality?) I've built MAT-based discrete VCAs and I've not seen 
     any *big* differences. Now keep in mind here that I'm talking about 
     10vpp waveforms being amplitude controlled into the line level 
     output, not studio-quality compressors or gates - that's probably a 
     totally different story and one that I have no expertise in at all.

     5) Op-amps - not an expert here either, at least as far as pro audio 
     goes. I can only say that the AD and LT devices have the best 
     reputation far as I've heard. For you chip smokers out there (i.e. 
     budding hobbyists), learn what you can using TL082s and you'll do 
     fine for now. Once you get a gut feel for things (which is what 
     happens after a few years), you can move on to critical spec 
     decisions.

     6) Power supplies - yup, sure enough, what should be the simplest 
     part sometimes can be the hardest to perfect. Power electronics is an 
     artform. The 78xx/79xx devices present us with yet another compromise 
     - so easy, yet not quite good enough. But they're good enough for 
     bench projects. The 723/4195's are good for small systems. The 723 
     with pass transistors will get you into the multi-amp high-wattage 
     regions, but with a lot of complexity. By this time it becomes more 
     efficient just to go out and buy some Power-One supply (which will be 
     a 723/pass transistor design anyway!).

     7) Electronotes is like the "Encylopedia Brittanica" of synthesizer 
     circuits. Even if you ignore the schematics and read only the text 
     from the first issue to the latest, you'll see the entire evolution 
     of the synthesizer industry, pretty much.

     Okay somebody, compile all these responses! This is a good thread!

     - Gene
     gstopp at fibermux.com

     p.s. hard vs. soft sync descriptions must be in the archives, 
     somewhere...

______________________________ Reply Separator _________________________________
Subject: RE: Favorite designs
Author:  Paul Schreiber <synth1 at airmail.net> at ccrelayout 
Date:    4/17/97 9:27 AM

I'll take a stab.

1) For VCOs, really can't beat a CEM3340. Extremely well behaved, = 
excellent tempreature tracking. Designs using uA726s should be avoided = 
at all costs (can't get uA726s at all for <$200). Not real sure if the = 
necessary +3300ppm temperature resistors are still around. Probably best = 
bet for a new discrete design would be a VCDO. Although not necessarily = 
"DigiKey friendly", it skirts the temperature issues (and introduces a = 
few quirks of its own). It really depends on the cost vs quality. The = 
original Moog VCOs are VERY drifty. Any transistor, by default, is an = 
exponential generator. But the temperature dependance of the Ies term = 
(not the kT/q term, although this is a small contributor) makes any EE = 
cringe. Heating up the transistor (like the uA726 does) is not really = 
solving the problem.

2) For a VCF, no contest: Moog ladder filter. If you can't beat 'em, = 
join 'em. The CEM3328/3378 comes close, but really there is no = 
substitute. Now, the "downfall" of a Moog is just that: it's a certain = 
sound associated with synthesizers. ARPs sound like ARPs. Most Oberheims = 
sound the same. The particular "sound" of a synth is 95% in the filter = 
characteristics. I really encourage everyone get a copy of Wendy Carlos' = 
CD "The Secrets of Synthesis" (I found mine at www.cduniverse.com) and = 
listen VERY CLOSELY to what she is saying. The message is: A Moog is a = 
Moog is a Moog. And I beat that dog to death.

3) For envelope generators, really can't beat a CEM3310 + a 555 for = 
variable gate delay.

4) For VCAs, really can't beat a CEM3381/3382. The OTA approach = 
(CA3080/LM13000) are just too noisy. Cheap, but noisy. Why build a whole = 
modular and run it out of a 45dB SNR part??? The CEMs have 100dB SNR! At = 
least use MAT-04 matched transistors for the front end.

5) For op amps: use AD or Linear Tech parts. The TL07x parts are just = 
OK, there are MUCH nicer parts. Rule #1: ALWAYS use dual, NEVER use = 
quads. Why? All of the really cool parts are duals, and quads crosstalk = 
too much (If you really need a quad, use the AD713KN. Awesome!) For = 
duals, look at OP275 or LT1113CS8. Stay away from junk like 1458s. Think = 
of a TL072 as the BOTTOM rung of the food chain.

6) The single most overlooked component is the power supply. Remember, = 
ever the lowly LM7815 was not around for the first analog synths. For = 
really good designs, check out the Audio Amateur Magazine (603)924-9464. = 
Folks, this mag is REQUIRED reading for all DIYs. Since the somewhat = 
demise of Electronotes, this is really the only source of precision = 
analog design. Stereos and CD players share alot with synths. For a = 
really good design, the 3 terminal stuff is NO GOOD. Perhaps I can = 
address this later.

7) The Electronotes series is simply the best way to get all the world's = 
knowledge of synth design. Call Bernie Hutchins 9PM to 11PM Eastern time = 
at 607-273-8030. At the very least, get the "Preferred Circuits = 
Collection". Then, save your money for the $265 (I think) "full set" = 
which is about 2000 pages.

Please, no "your the CEM chip guy" flames. Yes, I am. But I'm also the = 
"former Moog designer with 20 years experience designing this stuff" = 
guy.

Paul Schreiber
Synthesis Technology

----------
From:  Steve Jones[SMTP:sj2393 at ansys.com] 
Sent:  Thursday, April 17, 1997 3:53 PM 
To:  synth-diy at horus.sara.nl
Subject:  Favorite designs

Foregoing the hotter flames, what are some of our favorite 
designs here for the various analog modules.  I know this 
is an incredibly general (and perhaps dangerous) question. 
But I'm currently at square one and I'm trying to plot
out a course of action.  Modularity and flexibility are 
definite criteria.  So is frugality.  My plan is to procede 
incrementally, maybe starting with a single VCO and adding 
pieces as I go (sound familiar to anybody).

I have been encouraged by some to leave the CEM chips to 
those that really need them.  That sounds reasonable to me; 
so what are some discrete designs?  Gimme shematics! 8^)

This is indeed a general question.  But since there is no 
FAQ file per se, it certainly is, if rudimentary, on topic 
(or nothing is).

One more, slightly more specific, question: what is the 
difference between hard and soft syncing (surely a FAQ 
candidate)?

thanks, you guys are so cordial to beginners like myself, 
steve