great....would be nice if you use something like falstad for this kind of examples... that way we can see the flow of current and voltage......awesome your work
The OTA is really just acting as a a switch, so other switching mechanisms can work. Here, Ian Fritz uses a DG431 analog switch chip: ijfritz.byethost4.com/sy_cir4.htm
I should say the differential pair of the OTA is acting as two switches; you still need a current mirror (here, the PNPs pair at the top) regardless of what switching mechanism you are using.
You're a transistor, lightning resistor. Conducting to the mother star, that's what you are I don't even really like 311 but apparently that number is OWNED by a band in my head now.😄
Excellent Lecture, Ive been reviving some of my old synths, got my Minimoog D sounding better than it ever has and now trying to bring back osc. 2 on my Oberheim SEM ... Does anybody out there have any notes on the Theory of operation on the Oberheim SEM ? It has a few odd "quirks" and it seems tike osc 2 is trying to work but not quite...
professor, would the LM311 be uni polar +12V 7 GND or bi-polar?, the generated wave is positive only. to bad my dealer doesn't have the Ca3096, but does have the CA3086, CA3083 and i still have a few CA3046. i see many people use the CA3046 (like Doepfer), but not the CA3086 .. what would the reason be? the parameters look equal. Renesas HFA3096 is almost like 9€ / piece. but i can't find a PNP transistor array from the old intersill / harris line up. ya, i have the BCM857 but being sot23-6 not sure if i can solder that with my 25€ soldering station. that 28K7 doesn't seam to be necessary.
Presumably the LM311 is run off of a full bipolar +15 V (for VCC) and -15 V (for VEE), but note it has a separate GND input pin which is what the output is connected to in one of the output states. Re: the various 3xxx chips, I don't think there's magic to it, I'm guessing it's whatever people had most available at the time. Sorry, I'm not sure what you mean by the 28K7 not being necessary. It seems needed to get the v_tri switch levels to match up with what the comparator is feedback back.
@@Lantertronics yeah, i knew that pin to GND is pin of the internal transistor on the output. but i think it would probably also work having the LM311 connected uni polar. just wanted to be sure and prevent soldering things twice and trowing away things (hate breadboards), will make it bi polar. pretty neat thing that LM311, other wise the output would be negative. i ran the circuit i the "Falstad" simulator ( with a simple current sink ) and didn't place the 28K7, because Jörgen Bergfors schematic doesn't have it either and it worked fine, but will add a footprint. it's a real bummer i can't get the CA3096, but my my supplier has all other Intersil / Harris. to have matched like i said, i have the BCM857 but am scared i might mes it up while soldering with a blunt point that's like 0.05 of an inch wide. maybe i should opt for Ian fritz transistor matching. to bad Jörgen didn't publish his schematic
@@AnalogDude_ A year late to the party, sorry. Given that the input common-mode range of the 311 does not go down to the negative rail, a bipolar supply is necessary in this case.
Saw to square is easy, as at any point in time, the amplitude of saw tells you what phase is it in at that point in time. Squarewave doesn't do that. It doesn't contain that information. Which makes square to saw more difficult and kind of annoying. But it is possible and you can even involve things which wouldn't be possible with triangle core, like for example phase-locked-loops, or an differentiator to detect the edges of square & then perhaps measure time between edges and using that information some way to create new shapes... Or the most obvious way would be feeding square into integrator (+using some smart DC blocking) to get a triangle, but that would have loudness dependent on the frequency. Or even timbre dependent on the frequency, if you allow the integrator to saturate/distort at lowest frequencies & it may be a good idea to do that a bit, as otherwise the loudness dependency could lead to noise noise issues (lets say that our range is 9octaves from 20Hz to 10KHz. Each octave will be 6dB quieter than the previous one, so if 20Hz will swing between +-5V then 10KHz will be -54dB quiter swinging only +-10mV which may be bit just few 10s of dB above the noise floor... bit too close.. But in the end, most of the methods (except some weird ones like PLLs) of square waveshaping will lead you to some kind of recreating a saw or triangle from the square and in such cases it's just practically easier to rather start with a saw or triangle in the first place. Also most of the square-cores have some point in the circuit, where saw of the same frequency occurs, and if you have both square and saw available anyway, then it's easier to use the saw, which is more versatile & easier to work with. Although starting with square and going the complicated way to saw or triangle may lead to some interesting audible artifacts, so don't get discouraged by the stuff above and try it if you feel like, and you may stumble upon something interesting.
Current mirror is really bad. I have made simular triangle wave generator, and in that schematics should be used cascode current mirror comprised with 4 p-n-p transistors. Also what do you mean by match them? For current mirror two params are important: first is Vbe and second is hFE. You buy 100 same transistor and choose for example 20 with same Vbe. After that from that 20 transistor you chose several with same hFE. Chosing transistor in that way gives you garantee that cascode current mirror will work propely.
Wonderful lecture series! I find these extremely useful for my analog Eurorack module design. Please keep it coming!
Your videos are truly a gift. Thank you for sharing these inspiring practical lectures
Another great video!! Thanks again
You are welcome!
very interesting, thank you very much
Great lesson! Thank you for taking the time to do this :)
That switch is pretty ingenious.
wicked!
You've got a good "point" there about the 4K7 thing.
great....would be nice if you use something like falstad for this kind of examples... that way we can see the flow of current and voltage......awesome your work
is there a way to have this triangle core working without the OTA thing??
The OTA is really just acting as a a switch, so other switching mechanisms can work. Here, Ian Fritz uses a DG431 analog switch chip: ijfritz.byethost4.com/sy_cir4.htm
I should say the differential pair of the OTA is acting as two switches; you still need a current mirror (here, the PNPs pair at the top) regardless of what switching mechanism you are using.
You're a transistor, lightning resistor. Conducting to the mother star, that's what you are
I don't even really like 311 but apparently that number is OWNED by a band in my head now.😄
Excellent Lecture, Ive been reviving some of my old synths, got my Minimoog D sounding better than it ever has and now trying to bring back osc. 2 on my Oberheim SEM ...
Does anybody out there have any notes on the Theory of operation on the Oberheim SEM ? It has a few odd "quirks" and it seems tike osc 2 is trying to work but not quite...
I thought this guy was in his 20s with that voice, imagine my shock
Heh, yeah, I'm not so young anymore... ;)
professor, would the LM311 be uni polar +12V 7 GND or bi-polar?, the generated wave is positive only.
to bad my dealer doesn't have the Ca3096, but does have the CA3086, CA3083 and i still have a few CA3046. i see many people use the CA3046 (like Doepfer), but not the CA3086 .. what would the reason be? the parameters look equal.
Renesas HFA3096 is almost like 9€ / piece.
but i can't find a PNP transistor array from the old intersill / harris line up. ya, i have the BCM857 but being sot23-6 not sure if i can solder that with my 25€ soldering station.
that 28K7 doesn't seam to be necessary.
Presumably the LM311 is run off of a full bipolar +15 V (for VCC) and -15 V (for VEE), but note it has a separate GND input pin which is what the output is connected to in one of the output states.
Re: the various 3xxx chips, I don't think there's magic to it, I'm guessing it's whatever people had most available at the time.
Sorry, I'm not sure what you mean by the 28K7 not being necessary. It seems needed to get the v_tri switch levels to match up with what the comparator is feedback back.
@@Lantertronics yeah, i knew that pin to GND is pin of the internal transistor on the output. but i think it would probably also work having the LM311 connected uni polar. just wanted to be sure and prevent soldering things twice and trowing away things (hate breadboards), will make it bi polar. pretty neat thing that LM311, other wise the output would be negative.
i ran the circuit i the "Falstad" simulator ( with a simple current sink ) and didn't place the 28K7, because Jörgen Bergfors schematic doesn't have it either and it worked fine, but will add a footprint.
it's a real bummer i can't get the CA3096, but my my supplier has all other Intersil / Harris. to have matched
like i said, i have the BCM857 but am scared i might mes it up while soldering with a blunt point that's like 0.05 of an inch wide. maybe i should opt for Ian fritz transistor matching. to bad Jörgen didn't publish his schematic
@@Lantertronics the Falstad simulator failed on the Buchla Sync circuit, on my prototype I'm gonna have 2 sync circuits.
@@AnalogDude_ A year late to the party, sorry.
Given that the input common-mode range of the 311 does not go down to the negative rail, a bipolar supply is necessary in this case.
Hi, Aaron!
I was wondering if there where any synths that use Square Cores in the VCO with effective waveshaping, do you know of any?
Saw to square is easy, as at any point in time, the amplitude of saw tells you what phase is it in at that point in time. Squarewave doesn't do that. It doesn't contain that information. Which makes square to saw more difficult and kind of annoying.
But it is possible and you can even involve things which wouldn't be possible with triangle core, like for example phase-locked-loops, or an differentiator to detect the edges of square & then perhaps measure time between edges and using that information some way to create new shapes...
Or the most obvious way would be feeding square into integrator (+using some smart DC blocking) to get a triangle, but that would have loudness dependent on the frequency. Or even timbre dependent on the frequency, if you allow the integrator to saturate/distort at lowest frequencies & it may be a good idea to do that a bit, as otherwise the loudness dependency could lead to noise noise issues (lets say that our range is 9octaves from 20Hz to 10KHz. Each octave will be 6dB quieter than the previous one, so if 20Hz will swing between +-5V then 10KHz will be -54dB quiter swinging only +-10mV which may be bit just few 10s of dB above the noise floor... bit too close..
But in the end, most of the methods (except some weird ones like PLLs) of square waveshaping will lead you to some kind of recreating a saw or triangle from the square and in such cases it's just practically easier to rather start with a saw or triangle in the first place. Also most of the square-cores have some point in the circuit, where saw of the same frequency occurs, and if you have both square and saw available anyway, then it's easier to use the saw, which is more versatile & easier to work with.
Although starting with square and going the complicated way to saw or triangle may lead to some interesting audible artifacts, so don't get discouraged by the stuff above and try it if you feel like, and you may stumble upon something interesting.
Current mirror is really bad. I have made simular triangle wave generator, and in that schematics should be used cascode current mirror comprised with 4 p-n-p transistors. Also what do you mean by match them? For current mirror two params are important: first is Vbe and second is hFE. You buy 100 same transistor and choose for example 20 with same Vbe. After that from that 20 transistor you chose several with same hFE. Chosing transistor in that way gives you garantee that cascode current mirror will work propely.
I'm just analyzing the Buchla schematic. Yeah, there are probably better mirrors.
259th like, yes!