The vacuum tube’s forgotten rival.

March 27, 2022 2:05 PM

Ken Shirriff writes about the magnetic amplifier for IEEE Spectrum's History of Technology series. [Possibly timewalled without an account.]
posted by eotvos (18 comments total)

Wow, thanks Eotvos, that was really interesting -- I just subscribed to the print edition of IEEE Spectrum magazine for more!
posted by pH Indicating Socks at 2:47 PM on March 27, 2022

That is totally amazing, and I don’t think I understand why saturation of the ferromagnetic core causes the AC current to be unimpeded thereafter. Is it that the changes in the current produce much smaller changes in the magnetic field after saturation, or do the changes in the impedance really drop out entirely?

To understand how it works, first consider a simple inductor, say, a wire coiled around an iron rod. Such an inductor will tend to block the flow of alternating current through the wire. That’s because when current flows, the coil creates an alternating magnetic field, concentrated in the iron rod. And that varying magnetic field induces voltages in the wire that act to oppose the alternating current that created the field in the first place.

If such an inductor carries a lot of current, the rod can reach a state called saturation, whereby the iron cannot become any more magnetized than it already is. When that happens, current passes through the coil virtually unimpeded. Saturation is usually undesirable, but the mag amp exploits this effect.

posted by jamjam at 3:36 PM on March 27, 2022

You’d be hard pressed to find a mag amp in electronic hardware produced today

There are definitely mag amps in production today! 1. I worked on one just a few years ago in a still-popular DC GFCI product. (It's a bloody clever design that uses it to achieve very good accuracy around zero; I can't take credit for thinking it up.) 2. a major class of mid-price TIG welder (see Miller EconoTig, now called the Mk 8, I think) is built around a very large magamp. They have rubbish small-current control but mid- and high-end are quite sensitive and robust.
posted by introp at 3:56 PM on March 27, 2022 [4 favorites]

The article is right to mention the Elliott 803, of which there is a working specimen at the National Museum of Computing in Bletchley. From what I can tell, Peter Onion is its most dedicated caretaker, and if you find him at the museum, there is a good chance that you can ask him to open the broad book of schematics and show you parts of how it works. The 803 is a bit-serial machine, and so you will see how the bits jump from magnetic donut to magnetic donut as they collide, interfere, and combine to do the various arithmetic operations that need doing.
posted by tss at 4:26 PM on March 27, 2022 [1 favorite]

You’d be hard pressed to find a mag amp in electronic hardware produced today

Isn't that what a Bob Carver magnetic field audio amplifier is?

("an electromagnetic device that amplifies electrical signals utilizing a transformer's core saturation principle, and the core non linear property. It consists of an iron core with two or more coils wound around it")

They are still in production (under a different name, he sold the company), and I'm not hard pressed to find one, I'm listening to it now.
posted by StickyCarpet at 5:45 PM on March 27, 2022 [3 favorites]

Carver's magnetic amplifiers are pretty traditional amplifiers in their outputs; the "magnetic amplifier" part comes from the feedback system chosen for the power supply. They're definitely not mag amps.
posted by introp at 6:12 PM on March 27, 2022 [3 favorites]

I'm disappointed that this isn't about turtleneck-wearing audiophiles insisting on the "warm sound" of their special amplifiers while dismissing tube amps as the tools of poseurs.
posted by goatdog at 7:48 PM on March 27, 2022 [6 favorites]

archived version
posted by Ten Cold Hot Dogs at 10:46 PM on March 27, 2022 [1 favorite]

If always wondered how magnetic core memory worked, now I can stop asking "magnetic cores, how do they work?"

>I'm disappointed that this isn't about turtleneck-wearing audiophiles insisting on the "warm sound" of their special amplifiers while dismissing tube amps as the tools of poseurs.

Hmm. The magnets give a definitive soundstage free of the quantum confusion that can be heard in valve and solid-state amplifiers.
posted by k3ninho at 11:39 PM on March 27, 2022 [3 favorites]

Before the switch to LEDs, Magnetic Amplifiers were used in most traffic control lights. They were responsible for the "soft" fade in and out of the lights when blinking.
posted by ArgentCorvid at 6:52 AM on March 28, 2022 [2 favorites]

The magnets give a definitive soundstage free of the quantum confusion that can be heard in valve and solid-state amplifiers.

This might be even better than my idea of selling superconducting speaker cable installations to audiophiles. (The real money is in cryogen delivery subscriptions.) I'm happy to write a genuinely true paper comparing the expected noise with plots that have conveniently scaled axes if you'll give me a cut when you design, build, and market them. If the results don't turn out in our favor, we may have to find another approach. Lifetime? Insensitivity to cosmic rays?

Now I'm wondering if active, local magnetic field compensators for vacuum tube audio amps would be marketable. Sometimes I wish I didn't have any ethical reservations.
posted by eotvos at 7:04 AM on March 28, 2022 [1 favorite]

I don’t think I understand why saturation of the ferromagnetic core causes the AC current to be unimpeded thereafter. Is it that the changes in the current produce much smaller changes in the magnetic field after saturation, or do the changes in the impedance really drop out entirely?

The former but, practically-speaking, the result is the latter: the magnetic domains of the core are as aligned as the material structure allows. More field doesn't yield more magnetization.

(The magnetizing field is the integral of the material permeability. The permeability approaches zero in a strong magnetic field so technically the integral does sliiiightly keep increasing, but by such a trivial amount that it can't easily be seen.)
posted by introp at 8:15 AM on March 28, 2022 [2 favorites]

I highly recommend the excellent @ferritecorememory Instagram account for anybody who is tickled by this!
posted by TheCoug at 9:10 AM on March 28, 2022 [1 favorite]

I worked on one just a few years ago in a still-popular DC GFCI product. (It's a bloody clever design that uses it to achieve very good accuracy around zero; I can't take credit for thinking it up.)

you can't just leave that there and not tell us more.
posted by Dr. Twist at 11:55 AM on March 28, 2022 [1 favorite]

The working details are protected and I can't post them here, but if you look up "flux gate magnetometer" you'll get the idea (which dates back to WW2):

1. Take a very high-susceptibility (χ) core and put a winding on it.
2. Drive the winding with an AC signal to alternately drive the core into and out of saturation.
3. Sense* the output current and look at how the core saturates. If there's no background magnetic field, the positive and negative halves of the saturation (think "sine wave") will match. If there's a background magnetic field, the core will go into saturation *sooner* in the direction aligned with that field, so there will be a net bias that you can sense (simplest way is by just integrating the current signal).
4. If you're calibrated* well, you can turn that measurement into real magnetic field units.

The WW2 era magnetometers used a second winding on the core to do current sensing but there are simpler ways to do it these days.

Basically, the non-linearity of the cores at their saturation points give you a huge gain. Small change in input == large change in output (saturation starts earlier or later, depending on orientation).

5. If you think of "background magnetic field" here as the net (Biot-Savart) magnetic field generated by two supposedly-matched DC conductors running in opposite directions through the core, you now have a way to sense just *how* different the current in those two conductors is.

* much of where the special sauce is: cheap sensing and accurate self-calibration.
posted by introp at 1:08 PM on March 28, 2022 [3 favorites]

OK, that's super cool.
posted by Dr. Twist at 2:09 PM on March 28, 2022

superconducting speaker cable installations

Nothing beats speaker cable made from a couple of garden hoses filled with mercury.
posted by HiroProtagonist at 8:10 PM on March 30, 2022 [1 favorite]

Anyone interested in some DIY magnetic amplifier hackery should look here and here.
posted by HiroProtagonist at 8:14 PM on March 30, 2022 [1 favorite]