I read about this a long time ago. Development in amplification didn't jump from vacuum tube then to the transistor, there was something in the middle......the magnetic amplifier.
WWII Germany perfected the device, and I read that one of their V-2 rockets crashed in England and didn't explode. When it was salvaged and analyzed the scientists there pieced together a magnetic amplifier used in its construction.
I have a couple of books on them, and I heard that it could be used to amplify sound. I did a quick search before making this thread and the subject is a little dead.
Just today I found this awesome site:
Homemade Magnetic Audio Amplifier.
http://sparkbangbuzz.com/mag-amp/mag-amp.htm
This is a tantalizing bit from the first site I listed site:
"The mag amp is true amplification without the use of tubes, transistors or IC's but it does require the use of an AC power source. While most ac oscillators require the use of transistors, this amp could conceivably run using an AC power signal from a carbon arc or maybe even a zinc oscillator or similar."
I wonder if the experts here could build on this and make something really well performing.
I hope somebody is interested in this stuff as much as I am, it looks neat.
WWII Germany perfected the device, and I read that one of their V-2 rockets crashed in England and didn't explode. When it was salvaged and analyzed the scientists there pieced together a magnetic amplifier used in its construction.
I have a couple of books on them, and I heard that it could be used to amplify sound. I did a quick search before making this thread and the subject is a little dead.
Just today I found this awesome site:
Homemade Magnetic Audio Amplifier.
http://sparkbangbuzz.com/mag-amp/mag-amp.htm
This is a tantalizing bit from the first site I listed site:
"The mag amp is true amplification without the use of tubes, transistors or IC's but it does require the use of an AC power source. While most ac oscillators require the use of transistors, this amp could conceivably run using an AC power signal from a carbon arc or maybe even a zinc oscillator or similar."
I wonder if the experts here could build on this and make something really well performing.
I hope somebody is interested in this stuff as much as I am, it looks neat.
Last edited:
I am currently moving forward with this with full force, and designing a MUCH more advanced audio amplifier based on this basic circuit and decades old principle.
It currently has a few more features and additions. I'll be prototyping it in the next few days and I possibly might share the final schematics and plans with the world.
It currently has a few more features and additions. I'll be prototyping it in the next few days and I possibly might share the final schematics and plans with the world.
I haven't tried a mag amp, but I love that site.
Homemade CRTs, flame triodes, TEA lasers, FETs etc etc etc.
That guy takes DIY to another level.
The second link showing the useful application of a magnetic amplifier, controlling hi currents. I don't think it is for audio.
I included the second link because it has a bunch of useful information on the technology and has a bit of an introduction.
Here is a video from the same guy. YouTube - Homemade magnetic amplifiers made from common materials.
Oddly enough none of his commercially produced audio amplifier designs were magnetic amplifiers. Not even in the power supply. Probably because it's much easier to make a good one of either for audio many other ways.
I've been working all ******** night on my special amplifier design.
It got alot of neat features which solve some problems and others which make it more pleasent to use. I can't say much more.
And I saw that article on a commercial mag amp. It is typical that they would build one and charge a retarded price for it. It totally defeats the ******** point! $18 grand, are you high!?
It got alot of neat features which solve some problems and others which make it more pleasent to use. I can't say much more.
And I saw that article on a commercial mag amp. It is typical that they would build one and charge a retarded price for it. It totally defeats the ******** point! $18 grand, are you high!?
I tried making one, but I probably used poor toroids (ferrite, but I think they turned out being medium or low permeability) and not enough turns.
There are two general approaches, and one characteristic of note.
The first approach uses the saturation property of the core. By biasing the core with a static field which is isolated from the AC excitation, one or the other path through the core can be saturated, controlling the average output current (and distorting it noticably). These saturable reactors often use a common mode / differential construction, where the bias flux flows the same direction in two seperate magnetic paths (a common mode bias), while the excitation is applied in a differential manner. Because the excitation cancels out [1], we don't see any AC in the bias winding.
There are two typical construction methods: one, using a common E type core (although usually with a thinner center leg and larger winding windows), the "reactor" windings are placed on the outer limbs, in phase, in series. When the bias winding is unbiased, the AC fields support each other and the inductivity is quite high. When biased, the core is driven into saturation and the windings are effectively seperated from each other and are now air cored. Thus, inductance is very small and current can be large. The other common method uses two toroid cores, where each reactor winding is placed on each core, then connected antiseries. The toroids are stacked and the bias winding is placed over both of them. Thus, common mode bias, differential mode AC.
[1] This is the "one characteristic of note": The excitation only cancels when the applied flux cancels. When the closer-to-saturation path becomes saturated, its flux slams to a halt, therefore applying more flux to the other winding. Flux is now unbalanced and EMF is induced in the bias winding. The immediate result of this is coupling; fortunately, we can make use of this, by causing the bias winding to look like an AC short circuit -- drive it with a voltage source, or bypass it with a sufficiently large capacitor. This shunts AC current through the bias winding, reducing the stray inductance and therefore improving the on/off ratio (i.e., it becomes more "conductive" (technically, susceptive), because the inductance can be even lower when fully saturated (i.e., lower than the two windings with air between). This only works very well when the air-core equivalents are well coupled to the bias winding, as is the case for the double stacked toroids.
The other general approach is actually being used, right under your nose, and you don't even realize it. This approach is based on the property of remenance.
Inside every ATX power supply, there is a magnetic amplifier. The power transformer is wound for 12 and 5V, but not 3.3. How do they make 3.3? They start by taking half the 5V winding and filtering that. That's a start, but obviously, half wave rectification gets you only 2.5V -- not enough. So they put in the other leg, but not all of it, since that would make 5V again. They use a series inductor. But not just any inductor -- this one has a square B-H curve, so when the core is magnetized (and it's damn well magnetized, since it's in series with a rectifier), it's saturated and maximum current flows. The important part comes when the diode turns off: because the core remains magnetized (typical square ferrites retain about 80-90% at ~0.3T, while Metglas and square orthonol retain more like 90-95% at ~1.1T), it will still be saturated by the next cycle, so full current can still flow. If you simply put a (linear, unbiased) inductor in series, you'd just be naively wasting AC, and that's no good. There's something about these inductors.
Now, when the diode is off, you are free to play with the inductor however you like. It still inducts a little, so when the diode turns off, it makes a little flyback pulse, short because little flux got into it in the first place. Well, if you stretch that pulse out, say by biasing a 2N3906 into it, you can suck more flux out of it, moving it further around the B-H curve. Now it will absorb exactly as much flux before delivering current. If the core is large enough and has enough turns that the AC input doesn't cover the entire B-H curve, then you can use this to "completely" cut off the AC power, for the price of just a little magnetizing current (the forward and bias currents should cancel out, so this output won't actually need a bleeder resistor to keep the voltage from running away!).
Tim
There are two general approaches, and one characteristic of note.
The first approach uses the saturation property of the core. By biasing the core with a static field which is isolated from the AC excitation, one or the other path through the core can be saturated, controlling the average output current (and distorting it noticably). These saturable reactors often use a common mode / differential construction, where the bias flux flows the same direction in two seperate magnetic paths (a common mode bias), while the excitation is applied in a differential manner. Because the excitation cancels out [1], we don't see any AC in the bias winding.
There are two typical construction methods: one, using a common E type core (although usually with a thinner center leg and larger winding windows), the "reactor" windings are placed on the outer limbs, in phase, in series. When the bias winding is unbiased, the AC fields support each other and the inductivity is quite high. When biased, the core is driven into saturation and the windings are effectively seperated from each other and are now air cored. Thus, inductance is very small and current can be large. The other common method uses two toroid cores, where each reactor winding is placed on each core, then connected antiseries. The toroids are stacked and the bias winding is placed over both of them. Thus, common mode bias, differential mode AC.
[1] This is the "one characteristic of note": The excitation only cancels when the applied flux cancels. When the closer-to-saturation path becomes saturated, its flux slams to a halt, therefore applying more flux to the other winding. Flux is now unbalanced and EMF is induced in the bias winding. The immediate result of this is coupling; fortunately, we can make use of this, by causing the bias winding to look like an AC short circuit -- drive it with a voltage source, or bypass it with a sufficiently large capacitor. This shunts AC current through the bias winding, reducing the stray inductance and therefore improving the on/off ratio (i.e., it becomes more "conductive" (technically, susceptive), because the inductance can be even lower when fully saturated (i.e., lower than the two windings with air between). This only works very well when the air-core equivalents are well coupled to the bias winding, as is the case for the double stacked toroids.
The other general approach is actually being used, right under your nose, and you don't even realize it. This approach is based on the property of remenance.
Inside every ATX power supply, there is a magnetic amplifier. The power transformer is wound for 12 and 5V, but not 3.3. How do they make 3.3? They start by taking half the 5V winding and filtering that. That's a start, but obviously, half wave rectification gets you only 2.5V -- not enough. So they put in the other leg, but not all of it, since that would make 5V again. They use a series inductor. But not just any inductor -- this one has a square B-H curve, so when the core is magnetized (and it's damn well magnetized, since it's in series with a rectifier), it's saturated and maximum current flows. The important part comes when the diode turns off: because the core remains magnetized (typical square ferrites retain about 80-90% at ~0.3T, while Metglas and square orthonol retain more like 90-95% at ~1.1T), it will still be saturated by the next cycle, so full current can still flow. If you simply put a (linear, unbiased) inductor in series, you'd just be naively wasting AC, and that's no good. There's something about these inductors.
Now, when the diode is off, you are free to play with the inductor however you like. It still inducts a little, so when the diode turns off, it makes a little flyback pulse, short because little flux got into it in the first place. Well, if you stretch that pulse out, say by biasing a 2N3906 into it, you can suck more flux out of it, moving it further around the B-H curve. Now it will absorb exactly as much flux before delivering current. If the core is large enough and has enough turns that the AC input doesn't cover the entire B-H curve, then you can use this to "completely" cut off the AC power, for the price of just a little magnetizing current (the forward and bias currents should cancel out, so this output won't actually need a bleeder resistor to keep the voltage from running away!).
Tim
Maybe if the power line came in at a few hundred kHz magnetic amps would be more popular, at least for utility grade applications. Unfortunately inductance can't modify a DC source, and 60Hz is way too slow for audio. By the time you cough up an AC signal to regulate with a mag amp, you may as well modulate it to begin with instead of doing all the magnetic wizardry. If you have to use power semiconductors to generate the power carrier signal you've already lost the incredible ruggedness that mag amps might still possess as the one reason to use them anymore, for anything.
Thanks! I've been trying to understand if 60hz could be used for audio.
My understanding is you could made HF AC from a low frequency 60hz source
using non solid state components. No transistors.