Grin, thanks. I guess that means you think I should do the math tonight :^)
I've got to read this too.
George H.
Only doubling current? I'm aware of an Inductoheat beast that used that network (driven by a current-fed inverter, populated with about a hundred IRFP460s in total) for a few 100kW at 400kHz. Typical Q 10-30, for annealing/heat treating small diameter wire in high speed wire lines. This was 80s tech.
It's an exact voltage-current transformation of the LLC resonant circuit, commonly used with *voltage-fed* inverters for resonant SMPS (and also induction heating).
In the most general, it's the merger of two L-match circuits, used in any number of permutations for RF matching and filtering, which have been known basically since CW RF existed (one can use constant-k and m-derived theory to work with such circuits).
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com "George Herold" wrote in message news:674f588c-89fc-41bd-ab27-d208e96325c5@googlegroups.com... > This is from following an EDN link*. > http://www.accelinstruments.com/Applications/WaveformAmp/Magnetic-Field-Generator.html > > I'd have my doubts it's a "new" resonant circuit. But I've never seen it > before. > We do a series resonance for coupling to NMR coils.. more current would > almost > always be nice. Has anyone done this? > > George H. > > * > http://www.edn.com/design/analog/4441240/High-Frequency-Helmholtz-Coils-Generate-Magnetic-Fields?
NMR/MRI is a bit lean these days - funding for the NIH and NSF is way down. Private funding isn't picking up the slack. Some research universities have been funding purchases internally but that's also not picking up much slack. All the remaining companies are public - return for the stockholders is the business goal. Agilent pulled out of the business and sent a bunch of people packing. A few were picked up by JEOL. The rest have saturated the supply of NMR engineers and scientists - need yer driveway plowed?. Bruker looked at the situation and decided to consolidate R&D and manufacturing in Europe where it's more expensive but much harder to reduce headcount.
If you had sent me a resume 4-5 years ago, it would likely have been possible to "create" a slot, give you a bench, a network analyzer, a mess of quality tools, a hotrod workstation - and let you grow into something useful.
-- Grizzly H.
Thanks Tim, I sorta had an epiphany* driving home. Seeing the current slosh in phase to both caps and then double into the inductor.. (on resonance) You're saying (and JL said the same above... but shorter.) that I can adjust the cap ratio, and get any sort of current boost.. paying the price of higher resistive impedance and lower resonate frequency. As I said I'm pretty much an idiot about L/C resonaces once you get past two elements.. (hey, with three active bits there are two normal modes, resonances, as you increase the current boost ratio do the two resonances move closer together?)
... I should be able to make some mechanical analogy.. Driving two springs with a mass hung in the middle.. (there's no gravity in this model.)
Thanks aga> Only doubling current? I'm aware of an Inductoheat beast that used that
Grin.. no worries.. I was at a physics lab guy conference last summer.. hanging out at the bar one night* I was talking with a women who was in some NMR part of Varian. now laid off and doing lab guy stuff at a Uni. It was a bit weird, 'cause I known "physcis" nmr (some) and she did "chemist" nmr, 1/2 the time I didn't know what she was talking about. Fortunately there was a prof. from Wash. U. who could translate and we all had a wonderful time.
George H.
*the conference took over a bar for at least part of one night, there was a free food buffet and you bought your own drinks. I enjoyed it.
The coil has X&R, the cap has opposite X, partly cancelling coil X, so a given drive V gets you more i. To achieve that the L sees higher V than the amp delivers. Quite simple stuff.
NT
This is standard antenna impedance matching.
I notice that no authorship is claimed.
RL
On a sunny day (Thu, 21 Jan 2016 10:49:34 -0800 (PST)) it happened George Herold wrote in :
Where did you get 50 Ohms from?
Let's say you have a serial LC, and want to drive it. And lets say it is in resonance. Then, for the 'perfect' L and C, the impedance will be very low (>zero).
Say you have an amplifier with 100 Vpp 1 A amp output, now you can connect that to that LC with a 100 Ohm series resistor. that will lower the Q and limit the current to 1 App. Or you could use a step down transformer, if it is 10:1 turns ratio, then you get 10 Vpp at 10 A and can drive the LC via an 1 Ohm resistor with better Q.
Using a capacitive tap on the LC is similar to a physical tap on the coil.
You are familiar with this oscillator circuit, or should be: + | |--- --------->| | | |--- ( === C2 | ( | | ( L1 |---------| ( | | ( === C1 [ ] | | | /// /// ///
Here the opposite is happening, the feedback voltage from the source is transformed up, the current lower, the voltage higher. My favorite oscillator :-) with C2 = 1/2 C1. The same works like this:
Transmitter output stages.. same
On a sunny day (Thu, 21 Jan 2016 13:17:25 -0800 (PST)) it happened snipped-for-privacy@gmail.com wrote in :
100% agree.PS: + | |--- ------------>| | | |--- | ( | | ( | | ( ---===---| === (__| | | ( | | | ( ---||---- | | Cd /// ///
cd is BIG, just decoupling here, not part of the tuning.
I should be a little more careful; the exact equivalence I was thinking of, is shown here:
Note that L C, V I, and series parallel.
The article (Fig. 1) shows a series capacitor to a current source, which is nearly meaningless: unless the load is operated in an inductive frequency range (below resonance), it's simply adding voltage drop to the network, and certainly not changing the current. Current in the tank is still dependent on its properties (Q multiplication and all that), but the extra capacitor is rather superfluous.
(The article sinks lower, where it shows a TS250 amplifier driving this network, which I guess can be wired for voltage /or/ current mode output? In Fig. 4, they show a /voltage source/ instead, which has very different consequenes for the circuit's behavior! The TS250 amplifier, for all I can find about it, seems to be nothing more than a (CV output) audio amplifier with a stupidly low input impedance and some monitoring circuits (meters and fault protection). They say it can "amplify voltage, current and power", but don't provide any clue how to implement the other two, or what the performance specs are under those conditions!)
Anyway, back to the network. Compare this one, to the case of a series resonant tank, driven by a constant voltage source (VSRC, R, L and C all in series). Now put an inductor in parallel with the voltage source. You're drawing some inductive current, but unless the load is capacitive (below resonance), you aren't even reducing the current draw, and certainy not changing the voltage. (Voltage ratio is still simply whatever it is, from the RLC parameters and frequency.)
Which is a pretty standard situation: you'd typically drive a series resonant tank with a step-down transformer, in order to match to the incredibly low load resistance (in the 50mohm ballpark for induction heaters). The transformer inevitably has inductance, but it doesn't matter to the circuit.
However, if an inductor is placed in series with a parallel resonant tank, or a capacitor in parallel with a series resonant tank (as shown in my diagram), you get a second stage of matching.
You can decompose the LLC network into an LC series resonant, with a parallel resonant across the C leg. The explicit C (C18 in my diagram) is split into Cser and Cpar parts. This is actually an L-match network, which allows an impedance step-up from the voltage source. The parallel resonant half, too, is an L-match network, this time stepping down from the tank's impedance, to the ESR (R20 in my diagram).
Everything is flipped for the current-analogy version, of course: C9 and part of L3 act to reduce the source impedance, then the rest of L3, and C17, are series resonant and reduce that impedance to whatever R13 is equivalent to. (Of course the loss is usually on the inductor, but that's not a big deal, loss is loss. At least for modest Q. Things get kind of sloppy at low Q. But you probably don't need circuits like these for low Q, either.)
More generally, LC networks are necessarily non-dissipative, and this has direct consequences for the reflected source impedance. The network will always present an input impedance (as seen by the source, whether V or I) which is either proportional or inverse to the load resistance. You can add or subtract reactance to push a range of cases into one quadrant or another (namely, it would be nice to have an always-inductive load for voltage-source inverters, or capacitive for current-source), but there is no general solution: there will always be a load resistance where the input has a disastrously low or high impedance, or crosses through quadrants.
Which leads to the Smith chart. If you simply plot y = Imaginary(Z) versus x = Real(Z) for a given network (while varying frequency to generate the curve), you will get some kind of line, curve or circle. That's all it is, a handy way to draw complex impedance.
Consider the series resonant circuit: at very low frequencies, it has high impedance (capacitive), so the position is very far down on the graph (negative y). At high frequencies, it's inductive, and far up on the graph (positive y). At the resonant frequency, it has resistance (positive x -- passive networks will always be drawn on the right half plane. -x would be negative resistance, obtaining power from the load: you can get coordinates on the left half plane where amplifiers are involved!). What curve is this? A straight line, actually: Z = R + jX, and R is constant while X is varying positive and negative. So it's a line at (x = R, y = -inf to +inf). But more useful, as you will see: it's also a circle with infinite radius (the reals form a closed ring, meaning, -inf wraps around to +inf, so it is technically a closed loop, and not a line segment or anything.)
If we supply this series resonant circuit with a constant current I=1, then this graph also plots the voltage developed at its input terminals. The magnitude (distance from (0,0) to a point on the line) is minimum at resonance, or for a voltage source, the current is maximum.
If you instead look at the admittance Y = 1/Z of this network, then it traces out a closed circle, of diameter 1/R, center at (1/2R, 0), so it's tangent to the origin. At resonance, it's only conductive (Y = 1/R), and goes inductive (positive y values) or capacitive (negative y) around resonance. The min/max susceptance (B, where Y = G + jB) is at G = 1/2R and B = +/- 1/2R. At frequencies very far from resonance, |Y| is small (near the origin), which means little current flow from a voltage source (or a large voltage drop from a current source). Note that, as the circle approaches the origin, its tangent line becomes closer and closer to parallel with the vertical axis, which means the real component goes to zero faster than the imaginary part. (Of course, 1/0 is equivalent to +/-inf, so we've wrapped around the circle at this point.)
If you add a capacitor in parallel to that conductance, you introduce more reactance, which pushes that circle in the -y direction. But the diameter is still dependent on R, so you can't make an arbitrary restriction (like wanting B < 0) for all R.
Some practical consequences of this:
- Resonant induction heaters always need tuning. You can expand the allowable range, say by using a beefier inverter that can handle more reactive power, but there is no general "tapless" or "tunerless" solution possible.
- Antennas always need to be tuned. Large antennas can work with the fields to give good matches over a fairly wide range (a decade or two), but any finite structure necessarily has to have an upper and lower frequency limit, and that consequently puts reactance on its input in those frequency ranges. Antennas of simple design (like a good old dipole or loop) have an even narrower range where they are effective, and need that much more tuning. The LC network required for matching need no more than 4 adjustments (R, X seen by transmitter; R, X seen by antenna). The bandwidth seen by the transmitter (for a given tuner setting) will necessarily* be equal to, or lesser than, that of the bare antenna.
- Switching power supplies can't be designed for terribly wide ranges of V, I, because the load is reflected through the transformation network, and all the same rules apply. I mean, sure, but subject to the same practical limitations (extra inverter VA?).
*I'm not certain of this necessity. But it seems very likely. I'm not sure how I'd go about proving or disproving it...Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Grin.. OK I know 104 on a cap is 0.1uF, but less about resonators and coupling.
George H.
Just my confusion. We use a similar looking circuit for impedance matching.
Yeah I think I understand that now.. of course I'd really have to build something for it to sink in.
No I don't know that. I have only made one RF LC oscillator.. and mostly I just copied. In the above, if it's got voltage gain do you have to worry about driving the gate/ base to high above the source/collector?
OK Thanks for the words and nice ascii art.
George H.
I think fig 1. is just to show that there is twice the current in the coil and not that they are driving it with a current source. I think they are always driving this with a voltage source.... (aside: I never do that.. all "my" Helmholtz coils get driving by current sources.)
Yeah sorry.. it's pretty obvious that the edn article was written as advertising copy for their amplifier.
Uugh, Tim, that was a lot to swallow... I was OK with the smith charts till you changed it to admittance.. which I've seen in the past, but don't use. I know it helps you RF types to think in terms of admittance and not impedance... but it caused me a bit of a brain freeze.
As I said somewhere.. I think to really understand I'd have to build one.
Thanks again, george H.
On a sunny day (Fri, 22 Jan 2016 07:09:46 -0800 (PST)) it happened George Herold wrote in :
No, there is some gate current on the top of the LC voltage, but that will rise the source:
The JFET circuit will oscillate from a few mV supply upwards.
You mentioned driving your coils with a current source, I did huge coils (around a sports hall) to do simultaneous translation in the 50kHz to 150 kHz band long ago. In resonance for each channel, and to get enough bandwidth, channels had series resistors, and some used current sense feedback to 'make a current source' although they were driven by low output amps, Just like the audio guys use feedback to get zero output impedance, you can use a series resistor to sense current and make a current source. Anyways, things become simpler in resonance, Do you drive those coils in resonance?
Yea, amazing... ASCII art still works :-)
Fun, I think I remember that... you've got a cockroft-walton multiplier stage in there... (I didn't pay attention to how you did the oscillator.)
Yeah that's it.. Opamp-> pass transistor-> load-> current sense R->FB.
No resonance drive... mostly just DC. But I did some B-field reversal experiments with our optical pumping..
15 cm (6") radius HH coils, but not many turns. You flip the B field and if you flip it slow enough the spins can "follow the field".. And if you go fast they can't and get stuck. What is fast or slow depends on the precession frequency.. which you can control with how much residual field you leave as you go through zero. (you are flipping the B field along the z-axis, say, the residual field is what is in the x or y direction.)Anyway there was no problem flipping it fast.. but probably only ~10kHz or something... it was really more about the slew rate.
George H.
On a sunny day (Fri, 22 Jan 2016 09:42:02 -0800 (PST)) it happened George Herold wrote in :
Nope, just some JFETs in parallel. Not that that is really needed. But the LED powered by the gate current is the interesting part :-( In the old tube days you had grid current, I really like JFETs, reminds me of tubes. So as the gain is > 1, the voltage on the LC keeps rising and rising,
OK, I have seen that demo, maybe it was you pointed to the link some time ago. related to discussion how to detect that missing plane... (some company from down under). On the missing plane some German university says it is much more north closer to the equator. I had the same idea, that it landed on water and kept transmitting.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.