EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Artlav on May 11, 2015, 09:51:45 pm
-
Let's say you want to pump up an MRI magnet.
Or something else that requires not only hundreds of amps (500-1000A), but precise amount of hundreds of amps into an inductive load.
How can it be done?
I'm thinking of some sort of a switching regulator, naturally.
Back when i made a coilgun, i used something like this:
(http://i.imgur.com/LSBYrdI.png)
It could regulate the current in the coil at up to 200A.
Problem is, it only worked on the scale of a few milliseconds, and the diodes were already warm after one shot.
And i'm thinking of something that would run for seconds or even minutes.
So, the question is - is making a higher current supply simply a matter of better heat dissipation, or are there tricks and alternative topologies to be used?
If the former, then diodes are the hot and hard to find problem - anything in that range will dissipate about a KW of heat each, ones over 300-400A are difficult to find, and they can't be paralleled. How to get around that?
-
Hockey puck rectifiers (http://www.ebay.com/itm/Westinghouse-Rectifier-R7S00816-800-volts-1600-Amps-Hockey-puck-/111659590751?pt=LH_DefaultDomain_0&hash=item19ff6e245f) which can handle those currents are easy enough to find. You could use synchronous rectification (diode handles current for a bit, then a parallel IGBT takes over) but it's more complex to drive and probably more expensive ... you're going to need some massive cooling regardless.
-
Which magnet are you curious about? The field magnet or the gradient magnets?
The MRI Main magnet is charged once by a brute force DC current regulator once it has been cooled. A short link that is not superconducting during charging is across the power supply. There is a heater around the link, and it is then a ~100 Ohm load across the coil. When the proper current level is stored in the magnet, the heaters are switched off. The superconducting shorting link is rapidly chilled, forming a closed loop that contains the current. The charging supply is then disconnected.
This is known as a "Persistent Current Magnet" .
The supply is ramped during installation over 24 hours to 500-1000 Amperes, the current ramping being very slow to avoid thermal driven quenching. Quenching is bad as you loose all the liquid helium. Its also VERY loud and activates a blow out panel on the roof. Not that I would know, I was only working on a CT less then 200 feet away...
The gradient coils are the coils driven by amplifiers. They use a series of stacked output transformers in series driven by water cooled IGBT bridges. Typical might be 450-1000 amps at 750 V. The stacked transformers will be driven by different modes of PWM. The center bridge will be driven by a high frequency PWM for fine tuning and high frequency sweeps, the others will be low frequency transformers for brute force at low frequency.
Take a look at Powerex CM600HU24H IGBTs which were typical of a older Gradient Amp.
Steve
-
You need as low a voltage drop as possible, to maintain field. Dumping it back into the supply every half cycle gets you a good falling dI/dt (as good as rising), but doesn't help any with maintaining the energy -- you're cycling a portion of it (the delta I) in and out of your supply as reactive power all the time!
I'd heard of a superconducting UPS; it had a nice big coil inside, charged to the tune of 1000A, and was maintained with an SCR normally. (For this to be effective, it would have to be ~1V forward drop (1kW!) bleeding it down with a time constant of hours or days, suggesting an inductance of ~100H.) When power is needed (to charge the inductor, or to deliver energy back), the SCR is commutated off, and the inductor connected to the supply rails (probably around 350 or 700V DC) momentarily. If a 10A load is required on the DC rails, the voltage can be maintained for a full delta-I of 990A, which is 99.99% of the contained energy.
I don't actually know if this was a real product, or just someone's dream. 1kW seems like an awful lot of dissipation to idle with. 100H seems remarkable, considering ITER's massive coils are only ~10H (I forget what exactly, but you can calculate based on the rated current and stored energy specs -- E = 0.5 * L * I^2). This thing wouldn't be much smaller than a room full of lead-acid, and would incur quite a lot of quiescent power (cooling plus losses), not to mention cost.
Tim
-
Your power supply must have a big enough reservoir cap to absorb such energy.
The supply will be a bank of supercaps, they can take as much power as they deliver. So, no chance of overcharging anything.
Anyway, an easy thing to forget about - thanks for noting it.
The control can be described as this (Q1 is the high side switch, Q2 is the low side switch, which is parallel connected to the coil):
Um, so i should add a third switch parallel to the coil (Q2), and the existing ones are the Q1?
Not quite follow your idea.
Which magnet are you curious about?
I've been contemplating making an MRI-type machine.
Not human sized, something much smaller and likely for an element that is easier to detect than hydrogen - i'm mostly interested in the technology than in anything practically useful.
So, i want to be able to control the current in the main (resistive) magnet for a start.
You need as low a voltage drop as possible, to maintain field. Dumping it back into the supply every half cycle gets you a good falling dI/dt (as good as rising), but doesn't help any with maintaining the energy -- you're cycling a portion of it (the delta I) in and out of your supply as reactive power all the time!
Fair point, that circuit was designed for a coilgun, where fall time was as important as rise time.
Actually, now i don't think the energy recovery is even necessary - the resistive losses will dominate any saving achieved.
Which makes blueskull's idea all that much obvious (except for the dumping energy back part) - just a switch to charge/top-off the magnet, a switch to maintain the current, and a few protection diodes for the dead time.
I.e. something like this?
(http://i.imgur.com/F1D9A4l.png)
-
MPI supplies can supply that easily.
http://www.magnaflux.com/Products/MagneticParticleInspection/Equipment/MobilePowerPacks/tabid/261/Default.aspx (http://www.magnaflux.com/Products/MagneticParticleInspection/Equipment/MobilePowerPacks/tabid/261/Default.aspx)
Dunno how precisely you can set it.
-
Inductively sum the output of a bank of interleaved flyback regulators? You could have a low turns ratio on the output but distributing the current across the many parallel phases.
-
I have built previously a 1500A DC, 3phase, SCR controlled bridge for testing traction motors for locomotives(parallel hopkins). It had about 100v max voltage output. I know it's not directly applicable but what is your anticipated maximum output voltage required to overcome DC resistance? We also had a motor driven variac controlled field supply (3 phase but diodes) of about 1500A and 10volts max for the fields. It was very responsive, we had of course a step down transformer after the variac which drove the diodes.
-
Let's say you want to pump up an MRI magnet.
Or something else that requires not only hundreds of amps (500-1000A), but precise amount of hundreds of amps into an inductive load.
How can it be done?
I'm thinking of some sort of a switching regulator, naturally.
Back when i made a coilgun, i used something like this:
(http://i.imgur.com/LSBYrdI.png)
It could regulate the current in the coil at up to 200A.
Problem is, it only worked on the scale of a few milliseconds, and the diodes were already warm after one shot.
And i'm thinking of something that would run for seconds or even minutes.
So, the question is - is making a higher current supply simply a matter of better heat dissipation, or are there tricks and alternative topologies to be used?
If the former, then diodes are the hot and hard to find problem - anything in that range will dissipate about a KW of heat each, ones over 300-400A are difficult to find, and they can't be paralleled. How to get around that?
use FETs instead of diodes, look up synchronous buck converter
-
Your power supply must have a big enough reservoir cap to absorb such energy.
The supply will be a bank of supercaps, they can take as much power as they deliver. So, no chance of overcharging anything.
Anyway, an easy thing to forget about - thanks for noting it.
The control can be described as this (Q1 is the high side switch, Q2 is the low side switch, which is parallel connected to the coil):
Um, so i should add a third switch parallel to the coil (Q2), and the existing ones are the Q1?
Not quite follow your idea.
Which magnet are you curious about?
I've been contemplating making an MRI-type machine.
Not human sized, something much smaller and likely for an element that is easier to detect than hydrogen - i'm mostly interested in the technology than in anything practically useful.
So, i want to be able to control the current in the main (resistive) magnet for a start.
Hydrogen is the easiest one
why not use permanent magnets? easy to get much higher fields
-
Man, just the sheer amount of Nd needed to make a magnet that big... (or even more than 1T -- yes, higher can be achieved despite Bmax = 1.5T ish for NdFeB).
http://en.wikipedia.org/wiki/Halbach_array (http://en.wikipedia.org/wiki/Halbach_array)
It's possible, but man, the weight and expense would probably make such a monstrosity actually uneconomical compared to a superconducting version. That's saying something!
Not to mention the tempco. It certainly wouldn't change temperature quickly, but the nonuniformity and 1/f tail would go on forever.
Tim
-
If you don't need high voltages (<3V e.g), alternative way would may be a multiphase buck DC-DC working in current limit mode.
But those would not get very precise, likely in range 0.5-1%. You can get 400-600A pretty easily with these when properly designed.
-
Please tell me how can get one tesla from a permanent magnet?
Is a neutron star a permanent magnet?
-
Your power supply must have a big enough reservoir cap to absorb such energy.
The supply will be a bank of supercaps, they can take as much power as they deliver. So, no chance of overcharging anything.
Anyway, an easy thing to forget about - thanks for noting it.
The control can be described as this (Q1 is the high side switch, Q2 is the low side switch, which is parallel connected to the coil):
Um, so i should add a third switch parallel to the coil (Q2), and the existing ones are the Q1?
Not quite follow your idea.
Which magnet are you curious about?
I've been contemplating making an MRI-type machine.
Not human sized, something much smaller and likely for an element that is easier to detect than hydrogen - i'm mostly interested in the technology than in anything practically useful.
So, i want to be able to control the current in the main (resistive) magnet for a start.
Hydrogen is the easiest one
why not use permanent magnets? easy to get much higher fields
Please tell me how can get one tesla from a permanent magnet? If you managed to get a strong enough magnet to do MRI imaging, or even MRI chemical analysis, then please tell me, and we can make lots of money from that ^_^.
Don't forget you need a constant 1T field in a relatively large area.
There are parm magnet MRIs, but all are low resolution devices. Not practical for precision imaging or chemical analysis. At 60Mhz, you need 1T to get it resonant with proton.
Halbach array, we makes some ~200mm diameter, 25mm bore. With the highest strength magnetic material they are ~1.7T
but they stat to degrade about ~50'C, So high temp magnet are better and gets to about 1.5T. It is not high resolution because that isn't what it is used for but you can get high resolution permanent magnets, but they are expensive it takes a lot of tweaking to get them homogeneous
non-superconducting magnets just doesn't add up, high current or high voltage doesn't matter, they are huge, inefficient, need water cooling and or low duty cycles to get any where near 1T
H at 1T is ~42.5MHz
-
Man, just the sheer amount of Nd needed to make a magnet that big... (or even more than 1T -- yes, higher can be achieved despite Bmax = 1.5T ish for NdFeB).
http://en.wikipedia.org/wiki/Halbach_array (http://en.wikipedia.org/wiki/Halbach_array)
It's possible, but man, the weight and expense would probably make such a monstrosity actually uneconomical compared to a superconducting version. That's saying something!
Not to mention the tempco. It certainly wouldn't change temperature quickly, but the nonuniformity and 1/f tail would go on forever.
Tim
the tempco is a challenge, at around 1000ppm/C
I'd love to hear where you can get a superconducting magnet (and run it) for a less than x kilo of Neodymium magnets
-
Such a magnet will be dangerous, especially in a fault condition where excessive currents flow and therefore excessive magnetic fields are generated. Had a fault in a HV supply, 1500V, 3 phase scr, and the 5mH air cored filter inductor weighing 200kg moved several feet, crunching against the power transformer. Also have a friend that is an MRI technician and he has heard a fatalities of other technicians due to the inadvertent introduction of some tool effected by the magnetism.
-
Such a magnet will be dangerous, especially in a fault condition where excessive currents flow and therefore excessive magnetic fields are generated. Had a fault in a HV supply, 1500V, 3 phase scr, and the 5mH air cored filter inductor weighing 200kg moved several feet, crunching against the power transformer. Also have a friend that is an MRI technician and he has heard a fatalities of other technicians due to the inadvertent introduction of some tool effected by the magnetism.
This happend a few year ago when a mover thought he would just move the magnet out of the way
http://ing.dk/sites/ing/files/styles/w1120_media_right/public/images/54693.jpg?itok=F2H-VmyP (http://ing.dk/sites/ing/files/styles/w1120_media_right/public/images/54693.jpg?itok=F2H-VmyP)
that pallet jack is ~65kg, it too a few few weeks to plan how to get it off again
-
I'd love to hear where you can get a superconducting magnet (and run it) for a less than x kilo of Neodymium magnets
Any guesses how big an MRI sized 1T Halbach would be?
I have no idea what the price comparison actually is... and when you buy an MRI, you're buying a whole hell of a lot more than a magnet, so the cost of a PM alone might not compare, but that's not apples to apples either.
Tim
-
Such a magnet will be dangerous, especially in a fault condition where excessive currents flow and therefore excessive magnetic fields are generated. Had a fault in a HV supply, 1500V, 3 phase scr, and the 5mH air cored filter inductor weighing 200kg moved several feet, crunching against the power transformer. Also have a friend that is an MRI technician and he has heard a fatalities of other technicians due to the inadvertent introduction of some tool effected by the magnetism.
This happend a few year ago when a mover thought he would just move the magnet out of the way
http://ing.dk/sites/ing/files/styles/w1120_media_right/public/images/54693.jpg?itok=F2H-VmyP (http://ing.dk/sites/ing/files/styles/w1120_media_right/public/images/54693.jpg?itok=F2H-VmyP)
Nice photo. Those unconstrained fields are really strong.
that pallet jack is ~65kg, it too a few few weeks to plan how to get it off again