On 2023-03-08 12:59, John Larkin wrote:
> On Wed, 8 Mar 2023 11:23:50 -0500, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>
>> On 2023-03-08 10:47, John Larkin wrote:
>>> On Tue, 7 Mar 2023 22:15:32 -0500, Phil Hobbs
>>> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>>>
>>>>
>>>> A bunch of my stuff wouldn't budge off the peg with those
>>>> settings. If you already know what it should look like, why
>>>> simulate?
>>>
>>> To play with ideas fast day or night, and to get the parts
>>> values right.
>>
>> I've been known to dork frequency compensation that way too, and to
>> look at tolerances for filters.
>
> Right. Any interesting control loop is nonlinear.
>
>>
>> The last thing I designed by poking at in SPICE was a collapsible
>> BJT cascode, where the top transistor had to take a bunch of
>> voltage. When the bottom transistor turned on, the top one
>> saturated, pulling down its relatively high-Z base bias string, so
>> the low level was just the sum of the V_CE(sat)s.
>>
>> I knew that part would work, but needed to figure out what the
>> turn-off transient would look like. The answer was, "Nothing too
>> pretty, but good enough for government work." It didn't look like
>> the top transistor would reliably come out of saturation fast
>> enough to avoid blowing up the bottom one, but a zener on its base
>> solved that.
>>
>>> It's often easier to sim, say, a voltage divider than to do a lot
>>> of math.
>>
>> Our ideas of "a lot of math" are a bit different, but doing that is
>> no worse than using a calculator. ;) (I try to keep in practice
>> doing mental arithmetic, because it makes me a much better
>> troubleshooter.)
>
> A non-trivial voltage divider, with more than two resistors, can be
> a pain to work out with a calculator, especially using the parts you
> already have in stock.
>
> And Spice sims, with comments, are easy to save in a design notes
> folder.
Yup. They go in our git repos too (one repo per project).
>
> I do a lot of mental arithmetic, but it's analog computing,
> guesswork and not manipulating digits in my head. I usually come in
> around 5 or 10% accuracy, which is handy enough in most cases.
I like to make it a game. It's memory more than anything--in 1 Hz, 3 uA
of photocurrent has 1 pA shot noise; for Johnson, 60.4 ohms has 1 nV and
16k has 1 pA.
Going up and down decibels is fun too: 1 dB is 12% voltage and 25%
power, pi = 10 dB (voltage), and so 2 pi = 16 dB. A noise contribution
that's 6 dB down raises the floor by 1 dB, and it goes down linearly (in
dB) for lower noise.
> Doing this at a whiteboard, in a meeting, seems to annoy some people.
> This used to be called "Lightning Empiricism"; see Williams 1991.
Whiteboards are great fun, when combined with an interesting problem and
one-to-three smart and good-natured colleagues. (And possibly a pitcher
of beer.)
>>> I've invented some cool but improbable circuits by fiddling in
>>> Spice. My ac/dc programmable dummy load was an unexpected
>>> accident and is now a product.
>>>
>>> Doing the .opt tricks is safe if you are careful, which isn't
>>> hard. Run a slow sim, note a few points, booger the sim until it
>>> changes much, back off. I suppose that the options affect the
>>> internals of imported models, so all you can do is try.
>>
>> I've done that in a pinch. I normally don't simulate big wodges
>> of circuitry, so it doesn't come up much.
>>
>> Once in a great while I've needed to simulate the start-up or
>> ring-down of an oscillator, which can take forever, but you do more
>> of those, I expect.
>>
>>>
>>>>
>>>>>
>>>>> I've done sims that take an hour to run a few milliseconds.
>>>>> Feedback that slow doesn't train your instincts.
>>>>
>>>> Using universal opamps and VCVSes helps sometimes.
>>>
>>> I most always use one of the universal opamps, because LT Spice
>>> seldom models mine, and it's too hard to research their list to
>>> find one that's close to what I intend to use.
>>
>> Op amp models are such random-number generators anyway.... Once
>> in awhile you get one that models asymmetrical slewing, but not
>> very often.
>>
>> A depressing number even get the supply current all cattywampus.
>> Even UniversalOpamp2 gets that right!
>
> Thank goodness for that! I sometimes use the supply rails as the
> outputs of an opamp, to drive big power booster fets for example,
> and I need that to be right, statically and dynamically.
It's a good trick, for sure, but not all amps like it. Old-timey
'single supply' amps often have very very poor PSR on V-, because of
course the designer expects you to ground it, duh.
Since CMR is input-referred, you can get _voltage gain_ between V- and
the output, which makes life very exciting. UniversalOpamp2 will _not_
model that correctly. ;)
> As I noted recently, I found one ADI opamp that pulls 1e17 amps on
> V+.
Very non-green, for sure.
>
> Some will *output* free kilowatts from their power rails.
Still need a lot to make up for that 12V * 1e17 A.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com