7 ms·

Positive feedback loops are recursion, in many cases the skill in electronics engineering is avoiding positive feedback. Sometimes you can use a bit of positiv

Positive feedback loops are recursion, in many cases the skill in electronics engineering is avoiding positive feedback. Sometimes you can use a bit of positive feedback to suppress noise, as in a Schmidt trigger, which is bistable. Sometimes you can use it under manual control as in a Super-regenerative receiver. Mostly, it is used in oscillators to produce sine waves.

dreamcompiler 4y ago
In that sense negative feedback is recursion too because it's taking part of the output signal and feeding it back into the input again. Negative feedback is closer in spirit to a typical recursive function where you often want each iteration to process a "smaller" argument than the last and thus be closer to some fixed point.
- mikewarot 4y ago
  I do vaguely recall that the first time I saw an OP-amp circuit it seemed insane. It's only after you learn about negative feedback, and forcing a virtual ground that you can work with the things and not think they're black magic. That was decades ago, I'd forgotten about it.
ilyt 4y ago
> Mostly, it is used in oscillators to produce sine waves. Well technically to get sine you need a gain of exactly 1, anything above that will increase distortion. Now to get it to start oscillate in the first place you need above 1 so old school low distortion sine generators usually have some kind of automatic gain control
- mikewarot 4y ago
  Once upon a time, the #47 panel lamp served that function.
namibj 4y ago
Also useful to reduce drive power in hard-switched field-effect-device cascodes: with some mild tricks, stacked cascodes in particular benefit from compensating miller effect with positive feedback. It still all falls down again when the bottom node/switch is turned back on, and the only arguably difficult part is minimizing the extend of damping needed to prevent the positive feedback from oscillating, like it happily does with inductive load. The approach appears to be of relevance for MVDC power supply applications, as well-compensated capacitances in a stacked cascode can minimize driving/switching power in hard-switched SiC JFET based switches (at around 1~2 kV per JFET, stacked as high as needed), with usefully low loss up to the MHz range of e.g. PWM frequencies. Switched capacitance converters obviate the need for poorly-scaling higher-frequency magnetics, at the mild cost of needing some more switches (only divide-by-2 seems to be truly effective/ideal, while transformers can do arbitrary ratios), and the arguably massive gain of power density reachable in below-substation power classes (pole transformers, up to electric trains, the latter benefitting especially from skipping both the transformer and the AC filtering needs w.r.t. power-to-weight aspects).
mikewarot 4y ago
It took me a while to find it... but here it is, a motor generator that can give a gain of 10,000 or more using a bit of recursion https://en.wikipedia.org/wiki/Amplidyne https://en.wikipedia.org/wiki/Amplidyne ----- An earlier circuit that gets gain out of Motor Generators is the Ward-Leonard control https://en.wikipedia.org/wiki/Ward_Leonard_control https://en.wikipedia.org/wiki/Ward_Leonard_control At the US Steel works in Gary, Indiana, they have a 6 stand cold rolling mill. This uses pairs of 19,000 Horsepower motors on each stand to pull a cold slab of steel 6 inches thick like taffy to reduce its thickness. The electronics of the day (early 1960s) couldn't drive things directly, but they did have SCR packs big enough to drive the fields of DC generators in the Ward-Leonard configuration. I rebuilt many of those SCR packs in the 1980s.

dreamcompiler 4y ago

In that sense negative feedback is recursion too because it's taking part of the output signal and feeding it back into the input again. Negative feedback is closer in spirit to a typical recursive function where you often want each iteration to process a "smaller" argument than the last and thus be closer to some fixed point.

mikewarot 4y ago
I do vaguely recall that the first time I saw an OP-amp circuit it seemed insane. It's only after you learn about negative feedback, and forcing a virtual ground that you can work with the things and not think they're black magic. That was decades ago, I'd forgotten about it.

ilyt 4y ago

> Mostly, it is used in oscillators to produce sine waves. Well technically to get sine you need a gain of exactly 1, anything above that will increase distortion. Now to get it to start oscillate in the first place you need above 1 so old school low distortion sine generators usually have some kind of automatic gain control

mikewarot 4y ago
Once upon a time, the #47 panel lamp served that function.

namibj 4y ago

Also useful to reduce drive power in hard-switched field-effect-device cascodes: with some mild tricks, stacked cascodes in particular benefit from compensating miller effect with positive feedback. It still all falls down again when the bottom node/switch is turned back on, and the only arguably difficult part is minimizing the extend of damping needed to prevent the positive feedback from oscillating, like it happily does with inductive load. The approach appears to be of relevance for MVDC power supply applications, as well-compensated capacitances in a stacked cascode can minimize driving/switching power in hard-switched SiC JFET based switches (at around 1~2 kV per JFET, stacked as high as needed), with usefully low loss up to the MHz range of e.g. PWM frequencies. Switched capacitance converters obviate the need for poorly-scaling higher-frequency magnetics, at the mild cost of needing some more switches (only divide-by-2 seems to be truly effective/ideal, while transformers can do arbitrary ratios), and the arguably massive gain of power density reachable in below-substation power classes (pole transformers, up to electric trains, the latter benefitting especially from skipping both the transformer and the AC filtering needs w.r.t. power-to-weight aspects).

mikewarot 4y ago

It took me a while to find it... but here it is, a motor generator that can give a gain of 10,000 or more using a bit of recursion https://en.wikipedia.org/wiki/Amplidyne https://en.wikipedia.org/wiki/Amplidyne ----- An earlier circuit that gets gain out of Motor Generators is the Ward-Leonard control https://en.wikipedia.org/wiki/Ward_Leonard_control https://en.wikipedia.org/wiki/Ward_Leonard_control At the US Steel works in Gary, Indiana, they have a 6 stand cold rolling mill. This uses pairs of 19,000 Horsepower motors on each stand to pull a cold slab of steel 6 inches thick like taffy to reduce its thickness. The electronics of the day (early 1960s) couldn't drive things directly, but they did have SCR packs big enough to drive the fields of DC generators in the Ward-Leonard configuration. I rebuilt many of those SCR packs in the 1980s.