cr1901 changed the topic of ##yamahasynths to: Channel dedicated to questions and discussion of Yamaha FM Synthesizer internals and corresponding REing. Discussion of synthesis methods similar to the Yamaha line of chips, Sound Blasters + clones, PCM chips like RF5C68, and CD theory of operation are also on-topic. Channel logs: https://libera.irclog.whitequark.org/~h~yamahasynths
myon98 has joined ##yamahasynths
myon98 has quit [Quit: Bouncer maintainance...]
ejs has quit [Read error: Connection reset by peer]
<Sarayan>
andlabs: a visual6502-style netlist considers that the terminals have zero leakage, and the capacitors are only there to combat leakage
<Sarayan>
so you can ignore them and it will just work. It's *not* an analog circuit
<Sarayan>
it's a dynamic digital circuit, and that works without a hitch in software sims
myon98 has joined ##yamahasynths
<andlabs>
hmmm
<andlabs>
ok
<andlabs>
I still never fully understood how those netlists worked, honestly
<andlabs>
that sentence made it sound like it was using the capacitor for logic despite being digital
<andlabs>
so I still don't fully understand things all these years later
<andlabs>
granted I haven't actively tried learning eh
<Sarayan>
the structure of these rotating shift registers is inverter-clocked pass-inverted-inverse clocked pass-etc
<Sarayan>
so you can ignore the capacitors and understand how it works (the design is careful never to have clock and inverse clock asserted at the same time)
<Sarayan>
you have like 32 in a loop giving you 16 bits that rotate in 16 full clocks
<Sarayan>
with one place where you have an optional derivation for writing, and another where you read
<Sarayan>
and the whole design is synchronized on these cycles
<Sarayan>
the trick is in the real world when the pass is off the voltage on the gates is going to dissipate eventually (and relatively fast) even if in theory it wouldn't
<Sarayan>
so you add a (small!) capacitor on the gate so that it stays put for a clock cycle. That's dynamic logic, nmos era
<Sarayan>
which is also why you have a minimum clock rate for a bunch of circuits, if you go under it the charge dissipates
<Sarayan>
keeping a charge for an indefinite time requires two crossed inverters, so that's 4 transistors, takes a lot more space
<Sarayan>
that's called a static design, but the advantage is that the clock can be stopped or slowed down at any time and the state is not lost
<andlabs>
ah, neat
<andlabs>
I suppose then a netlist format that stored the information about capacitors owuld only be really useful if you're looking to actually physically recreate the chip, but one that only stored the digital logic ifnormation would be more useful for simulation
<Sarayan>
yep
<myon98>
I think I read at the visual6502 FAQ that, except for the cases of shorting two of them together, assuming that floating nets will keep their values indefinitely can be good enough for digital simulation purposes
<myon98>
I wonder how was it like to design chips with Rubylith tapes, or how much CADs were available at the time of DX7 or other Yamaha's NMOS FM chips
SceneCAT has joined ##yamahasynths
Stiletto has quit [Quit: leaving]
Stiletto has joined ##yamahasynths
<Sarayan>
myon98: 68000 has a lot of different transistor shapes, iirc, yamaha (2203 at least) it a little more regular. But that may just be a result of their heavily repeast design
<Sarayan>
repeated
ejs has joined ##yamahasynths
<cr1901>
Sarayan: By "zero leakage" do you mean "leakage due to current being sent through/from terminals is ignored"?
<cr1901>
Or "leakage due to inherent capacitance in the wires connecting to each terminal is ignored?"
<cr1901>
(Both are kinda weird to me because you have to treat the terminal input as holding a charge/voltage even when the mosfet is off)
<Sarayan>
leakage due to the fact that even when the gate is a gnd the isolation is not perfect so electrons (or holes) eventually go away and everything ends up at the same potential
<Sarayan>
technically there's even leakage from gate to terminals, depending on the quality of your isolation
<cr1901>
So to model leakage (simply), model the mosfet when gate is GND as a resistor?
<Sarayan>
possibly, I don't know the models of non-ideal mosfets
<cr1901>
I thought the capacitors where there between stages b/c even w/o leakage, you need a place to store the charge when the mosfet switch opens
<Sarayan>
if there wasn't leakage the charge wouldn't move
<Sarayan>
electrons don't just pop out :-)
<Sarayan>
(or in)
<Sarayan>
capacitors essentially add friction, they kinda move more slowly, so leak out slower
<Sarayan>
(10kft view, don't have a physicist hit me)
<cr1901>
Sarayan: I'm not having a good weekend. Idk where I was going w/ this.
<cr1901>
I think I was replacing the mostfet and inverter input w/ "an ideal switch and large but finite resistor"
<cr1901>
if you open an ideal switch, the voltage across the resistor will be 0 once it's opened
<cr1901>
w/ a capacitor, voltage across resistor will be nonzero for a time
<Sarayan>
there are parasitic capacities everywhere though
<cr1901>
Yea, but that's not gonna hold a charge for any meaningful time. Otherwise we could forego them in the inverter chains :D
<cr1901>
Reminds me of this presentation from 2014, where using parasitics to your advantage is a fairly novel thing in power electronics: https://www.youtube.com/watch?v=PoChrjIN84E
<NiGHTS>
David Perreault - Powerful Circuits: Developments in High Frequency Power Electronics - YouTube
<qu1j0t3>
cr1901: not sure it's entirely novel, some 60s circuits i've been looking at seem to rely on it
<qu1j0t3>
e.g. LC primary side oscillators i think
<qu1j0t3>
you can find transistors and transformer and no visible C
<cr1901>
How'd they reliably get a value for C in the 60s?
<cr1901>
(Anyways, you may be right, I was just reminded of the presentation)
<cr1901>
And Cover-My-Ass: "using parasitics to your advantage is a fairly novel thing in power electronics" is me paraphrasing what I got from the vid
<qu1j0t3>
:)
<qu1j0t3>
i think the circuit i am thinking of is probably from the GE Transistor Handbook 7th ed, i'll look it up some time
<qu1j0t3>
but actually i think it may also be a feature of the Tek 606 high voltage supply that i have been trying to look at
<cr1901>
You can use maxwell's equations and friends to find the capacitance, inductance, etc, of lots of stuff by hand.
<cr1901>
But my recollection is that it gets gnarly to do by hand if all 3 spatial dimensions are finite
<qu1j0t3>
i'm no expert so i could be wrong, but i don't see explicit C in the primary side oscillator https://imgur.com/gtO35ue
<NiGHTS>
Imgur: The magic of the Internet
<cr1901>
which is, well, real materials :P
<qu1j0t3>
this is actually not precisely the same circuit, but closely related, and same topology as my (dead) unit
<qu1j0t3>
there are lots of people here who could read this better than i can
<cr1901>
Perhaps post onto EE.SE and someone will give you an answer :P?
<qu1j0t3>
cr1901: idk, i don't have a very well framed question yet. i'm trying to figure out if the trnssformer is okay or not.
<andlabs>
also note that Yamaha did not manufacture their own components at the time of the DX7
<andlabs>
I don't know how much manufacturer difference there is when youg et down to the actual electronics
<andlabs>
but i have to imagine that they would have needed to translate designs between fabs somehow
<Sarayan>
cr1901: precise C was relatively easy, because it the surface times a constant. Precise r is very hard though. That's why the votrax uses switched capacitors to replace resistors in its filters
<Sarayan>
well, the easiest, which is what votrax uses, is precise C1/C2 ratio, which is pure surface ratio
<cr1901>
>because it the surface times a constant
<cr1901>
If the surface is infinitely long/wide :D. Idr how badly that falls apart for "real" surfaces
<cr1901>
(It might not :o. This is what I get for not keeping that knowledge ready in head.)
<qu1j0t3>
cr1901: "The capacitance is a function only of the geometry of the design (e.g. area of the plates and the distance between them) and the permittivity of the dielectric material between the plates of the capacitor"
<qu1j0t3>
cr1901: you can also just ... measure it.
<cr1901>
Measure it when you're designing a chip?
<qu1j0t3>
oh, on-chip ones.
<qu1j0t3>
well the same equation applies, afaik.
<qu1j0t3>
obviously in the olden days trimming and precision was more difficult.
<qu1j0t3>
i thought we were still talking about parasitics and power :)
<cr1901>
Tbf to you, I am going all over the place today
<cr1901>
I'm not _actively_/deliberately trying to move the goalposts this time :)
<qu1j0t3>
cr1901: it's fine, take it easy :) sorry to hear it's a bad weekend
<qu1j0t3>
the topic drift is my fault since the thread was originally about on chip caps
<Sarayan>
topic drift is one of the pleasures of life :-)
<cr1901>
Topic drift is a feature here in ##yamahasynths
* qu1j0t3
lies back in the flimsy raft, adrift on the sea of topic
<Sarayan>
better than a sea of gender, the drift there is due to rapids ;-)
<andlabs>
set adrift on (computer) memory bliss
endrift has joined ##yamahasynths
endrift|ZNC has quit [Ping timeout: 250 seconds]
<Lord_Nightmare>
ejs: "Victoreen VXR-2700" tube sounds like a tube used (maybe as a voltage regulator?) in an old victoreen geiger counter, possibly one of the civil defense ones
<Lord_Nightmare>
I have a working victoreen CDV-700 (which was last calibrated in the 1980s) and a very broken picker geiger counter which is really only good for parts
<Lord_Nightmare>
the CDV-700 is not fancy or extremely accurate unless kept religiously calibrated, but it is a very simple device and is pretty easy to keep 'working' although not terribly accurate
<Lord_Nightmare>
newer geiger counters are much more accurate