Of course, there's a reason why most of these aren't used anymore and even valves are used in very specific applications (or let's say, not all of them exactly for pure engineering reasons)

Semiconductors blow all of these out of the water. And as technology for higher power/higher frequency gets better they will be replaced more and more by semiconductors.

I absolutely love glasslinger's videos, thay're quite a gifted engineer.

Power control for serious power was really hard until about 30 years ago, when power MOSFETs and IGBTs got good. There were a long series of not very good devices from 1900 on. Ward-Leonard drives, dynamotors, magnetic amplifiers, motors with multiple switchable windings, thyatrons, ignitrons, silicon-controlled rectifiers... People needed smooth control of high power, but it wasn't easy to get.

Now you can get IGBT modules able to handle the power for a locomotive motor. They're not that big - you can hold one in your hand. The next thing seems to be going back to MOSFETs, but in silicon carbide. Tesla switched from IGBTs to SIC MOSFETs. At last, high power control is a solved problem.

Now locomotive power engineers can even have fun turning bugs (harmonics in the audio range) into features (solfege):

https://www.youtube.com/watch?v=_iLrXDU31Dg

For anyone else wondering what's going on there, certain types of semiconductors have a minimum off time. They have to work around this by varying the frequency of the PWM. Usually it results in a kind of space age sound as they accelerate. In this case they have chosen something more interesting.

yeah but what are the specific benefits of magnetic amplifiers that you cannot get from SICMOSFETs

Not much. There's a reason they're largely forgotten!

I guess there are fewer ways to destroy stacks of laminations and copper windings. But that doesn't make up for the extra weight, space, cost, and labor to manufacture them

You can build one at home?

Overload resistance and overall robustness.

A lifespan of a century or more?

I had the job of repairing some of these. They were made by Westinghouse in 1964 as part of a control system running an 6 stand rolling mill. The nice thing about them is you can have several inputs, all electrically isolated from each other (separate sense windings) and they worked all the way down to DC, which surprised the heck out of me.

The driver was a 10khz square wave, so the frequency response was DC to 1khz... fast enough for control loops.

Down to DC? That should be a surprise, because it's certainly not true! Faraday's law of induction only applies to varying fields, which means AC or physical movement. Remember, it's perfectly normal and common to have a square wave with a DC component of zero!

I think he means the response frequency, not the frequency of the AC drive (which he states was 10 kHz).

Yeah, they’re almost definitely referring to the control signal. Since it’s controlling the saturation of the core, it’ll happily go down to DC just fine.

Amazing, I had never heard of these things.

There is a 1960 Popular Electronics article about them: https://www.rfcafe.com/references/popular-electronics/magnet...

It's really the 1850's when you think about it.

Transformers were all there was before the Edison Effect was discovered and vacuum tubes were developed.

A lot can be accomplished with just copper and iron and its always been expensive and heavy, sometimes like a baseline for cost.

By comparison alternative systems can be much more complex with way more points of failure if they can just beat that cost to a certain extent.

There's more proven technology on the trailing edge than the leading edge can be expected to provide.

In one person's lifetime there always exists more forgotten technologies of the ages than truly novel technologies that can be developed over a single career.

It could be worthwhile to try and balance the creation of new technologies with the preservation of technologies at risk of loss.

Funnily enough, the "Edison" effect was originally discovered in the 1850s. It got rediscovered approximately every ten years after that.

Looks like Lars Lundahl (famous for high quality audio transformers) wanted to build an audio magnetic amplifier in the 90s, but the project got canned. Some German company ended up making it a decade or so later

http://www.auditorium-23.de/MagAmp/MagAmp.html

http://www.acousticplan.de/html/magamp_english.html

The technology has also been used in a completely passive multi-band eq/saturator - https://www.soundonsound.com/reviews/gyraf-audio-gyratec-xxi...

Saturable reactors are critical in semiconductor manufacturing. They are the key components in the magnetic pulse compression circuits that drive the excimer lasers used as light sources in photolithography and in the annealing of thin film LED displays.

Thanks for the comment, just learned something new today. I thought magnetic pulse compressors are only for physics experiments and EMP weapons, never knew they are used in photolithography.

Mass production. Thousands of units. When I learned this engineering, I briefly looked around for other contemporary applications. I couldn’t find a single one! Other than research and defense, pulse power is only applicable to driving pulsed lasers with very short upper state lifetimes. And magnetic pause compression, not being cheap, is only applicable in high repetition rate systems. Hence excimer lasers in the semiconductor industry.

Another lost amplification technology is the mechanical amplifier - basically a transducer which is mechanically coupled to a carbon button mic diaphragm - obviously only useful in the AF range, but it was how early telephone toll circuits were amplified and extended beyond what heavy gauge copper and loading coils could do on their own.

I've heard of these before; they are very cool. The only problem, as far as I've seen, is a smaller frequency range and lower gain compared to tubes or transistors. Not crazy smaller or lower, but don't expect a single magnetic ham radio amplifier to get you from 160m to 10m at 1.5kW legal limit.

Oh I recently came across a similar topology in welding machines. The Amp output was a large screw shifting a rod (shunt?) in the transformer. I had no idea it was a larger principle at play.

So if you wound a 1:1 transformer on a coil that has additional windings to ’preload’ it, could you use this as a speed controller/voltage regulator?

Yes, it's called a saturable reactor [0] and can be considered as a very simple type of magnetic amplifier. You control the impedance of the reactor using the extra winding, conceptually similar to how you control the conductivity of the pass transistor in a solid-state voltage regulator. You can find them in old Tektronix oscilloscopes, here's a long (70 min.) and comprehensive video [1] on the design of the Tektronix 555 power supply, including the operation of the saturable reactor voltage regulator.

[0] https://en.wikipedia.org/wiki/Saturable_reactor

[1] https://www.youtube.com/watch?v=CAbeWXcuYOE

Awesome thank you!!!

Yes. Saturable core reactors have a long history of industrial regulation of heavy duty motors, furnaces, and the like. I first learned about them from old cranes in seaports.

For some applications size and weight aren't a big deal, but durability and overload tolerance are, so I can imagine they are still used out there somewhere.

> EMP-proof

This detail caught my eye. Very cool.

But why? An EMP would still induce energy into the coils, largely disrupting operation, no?

But afterwards they'd be fine unless their isolation was compromised somehow through for instance arcing. And even in that case, there are ways to stop an arc and leave the system functional (self repair, essentially the coil material evaporates a bit further from the breach point than the isolation material).

In silicon if such an energy pulse would reach the circuitry it would be ruined forever.

They're still affected by EMP, but they're not damaged: as soon as the pulse is over, everything's back to working.

This is very much not true for modern VVLSI....

>"The mag amp can modulate, switch, invert, convert, multivibrate, audio-amplify, radio-amplify, frequency-shift, phase-shift, and multiply. Stages can be cascaded. Simple feedback techniques enable gains in the millions.

The mag amp can even compute. Trouble-proof magnetic binaries replaced the less reliable vacuum tubes used in some early digital computers."

This is my new favorite HN article in the field of electrical engineering...

Basically, what a transistor is to DC -- A Magnetic Amplifier is to AC... plus there is additional functionality...

This device is highly worthy of future study...

Here are some more schematics for audio amplification and receiving radio: http://www.aes.org/aeshc/pdf/how.the.aes.began/vincent_magne...

Im very curious what the tone of such amps is like. Anyone heard one in person?

Interesting, this seems like a practical implementation of the a-field (magnetic vector potential) "circuits" described here[1].

I wonder how hard it is to make an electro-electret...

[1] http://amasci.com/elect/mcoils.html

I love a good used book store. Two years ago I found this:

Storm, Herbert F, Magnetic Amplifiers, Wiley, 1955. LOC# 55-6432

Covers applications and theory, contains All The Math.

Looks hard to manufacture on scale for cheaper than a transistor amp. It was probably sidelined due to complexity of assembly.

Can this be used to control the massive magnets needed for nuclear fusion in tokomaks?

I've heard that some modern PC PSUs use these devices inside.

why did they fall out of favor.m?

Are they slow? Expensive? Something else?

I guess mostly price - cooper is more expensive, and the economy of scale in semi-conductor manufacturing should be much higher.

Expensive, heavy and they require a relatively large amount of power. Very, very robust though.

huh, I always wondered what those little loops of what looked like a "life saver" lolly was when I was pulling shit apart as a kid.

Those would most likely be inductors or transformers. It is mentioned in the article how the simplistic Figures 2 & 3 magnetic amps would act as a transformer. https://en.wikipedia.org/wiki/Toroidal_inductors_and_transfo...

How much EMI does it cause?

How long is a string? No more than any other magnetic component, such as the switching transformer in your cellphone charger...

The low-frequency operation means this will be quite low, and below the 250kHz lower threshold that most EMI testing cares about.

One of the most interesting articles I've ever read.

I did a fair amount of reading about vaccum tubes back in the day, and my family owned a tube-powered commercial Seeberg 45 RPM jukebox, but I never made the mental leap to all-magnetic "circuits."

This jukebox was in our living room (size of a refrigerator):

https://www.prosoundweb.com/put-another-nickel-in/

One disadvantage of magnetic amplifiers that I can think of is that any heavy-current application would require specialized training to service. For example, I work with 1-phase electricity as needed, but I won't even open a 3-phase panel. Can you imagine poking a screwdriver around in a magnetic circuit? :)

Fun fact: after the first vacuum tube amplifier was patented, patent trolls would solder random collections of vacuum tube logic (VTL) and patent those. But when cross-examined in court while defending their patents, they could not explain how they worked. Kind of like the first spaghetti code, but in hardware.

It would be very cool if somebody made a graphical simulator for magnetic circuits and amplifiers, like Kerbal Space Program (KSP.)

> Can you imagine poking a screwdriver around in a magnetic circuit? :)

I wouldn’t bet my life on it but my expectations is that you don’t feel much of the high frequency magnetic fields in your screwdriver.

Did you also know your Seeburg model used magnetic core memory to hold the music selections?

http://www.west.net/~jay/fmillera/fmillera/tormat.htm

I don't know about KSP simple, but Elmer is an GNU-licensed (iirc) open source multiphysical simulator, including magnetics.

https://www.csc.fi/web/elmer

Mag amps are still used for post-regulation in switching power supplies, such as the Mean Well QP-200 and some 80%+ efficient ATX power supplies.

+1.

See also: Magnetic Amplifier Control for Simple, Low-Cost, Secondary Regulation by Bob Mammano, Unitrode Corporation (now Texas Instruments).

https://www.ti.com/lit/ml/slup129/slup129.pdf

Are these still around though? Single-rail designs have been trickling down into lower and lower price points, meanwhile the cheapests power supplies never had independently regulated rails in the first place.

This could be really useful.

Reply on news.ycombinator.com