Hi!
I'm working in a french research institute, and one of our experiment uses RF coils to generate RF magnetic fields, from 100kHz to 1MHz.
Coils are grouped by 2 in parallel, and there are capacitors in series, in order to make a RLC resonant system to bring back the complex impedance to a real one, so the amplifier sees a pure resistor. This resulting resistor is around 1-2 Ohm. As the RF amplifier currently used (AG1010 from T&C Power Conversion, Inc.) needs a 50 Ohm load, there is also a impedance adapter, made with a simple tore transformer.
But we have troubles with the AG1010, wich don't like reverse power that sometimes appear during the resonant tuning phase.
An idea is to replace the AG1010 with Class D amplifiers. We don't need the analog to digital conversion, as we only need pure sinus: we can directly generate the PWM and drive the H-bridges. If I'm right, the class D is able to drive low impedance loads, so we can remove the impedance adapter.The needed power is less that 1kW per amplifier, and we need 4 amplifiers.
As we don't have the knowledge to design such amplifiers (it is simple in theory, but in practice, there are huge problems, like EMI...), we have 2 options:
1) find a company able to design this power stage for us;
2) find enough support to help use build it.
If you have some inputs for either item, I would appreciate.
Thanks for reading.
Frédéric
I'm working in a french research institute, and one of our experiment uses RF coils to generate RF magnetic fields, from 100kHz to 1MHz.
Coils are grouped by 2 in parallel, and there are capacitors in series, in order to make a RLC resonant system to bring back the complex impedance to a real one, so the amplifier sees a pure resistor. This resulting resistor is around 1-2 Ohm. As the RF amplifier currently used (AG1010 from T&C Power Conversion, Inc.) needs a 50 Ohm load, there is also a impedance adapter, made with a simple tore transformer.
But we have troubles with the AG1010, wich don't like reverse power that sometimes appear during the resonant tuning phase.
An idea is to replace the AG1010 with Class D amplifiers. We don't need the analog to digital conversion, as we only need pure sinus: we can directly generate the PWM and drive the H-bridges. If I'm right, the class D is able to drive low impedance loads, so we can remove the impedance adapter.The needed power is less that 1kW per amplifier, and we need 4 amplifiers.
As we don't have the knowledge to design such amplifiers (it is simple in theory, but in practice, there are huge problems, like EMI...), we have 2 options:
1) find a company able to design this power stage for us;
2) find enough support to help use build it.
If you have some inputs for either item, I would appreciate.
Thanks for reading.
Frédéric
It is not quite that simple, designing switching stages to deal with reactive loads, that can appear capacitive or inductive is not trivial, and a quick play with a smith chart and spice will reveal that you can easily get into SOA trouble even in a switching design.
Personally, I would skin this by keeping the existing amps and switching in a 10dB pad between the amp and the matching transformer during tuning, this will guarantee a 20dB return loss (VSWR ~1.2:1) at the amplifier during tuning which will keep it happy while you futz with the tuning capacitor, then switch the pad out for normal operation, there is little to be gained by tuning at full power, except accidental RF burns and arcing.
73 Dan.
Personally, I would skin this by keeping the existing amps and switching in a 10dB pad between the amp and the matching transformer during tuning, this will guarantee a 20dB return loss (VSWR ~1.2:1) at the amplifier during tuning which will keep it happy while you futz with the tuning capacitor, then switch the pad out for normal operation, there is little to be gained by tuning at full power, except accidental RF burns and arcing.
73 Dan.
You exceed the SOA. The SOA is the area of the transistor operation where currents and voltages are OK so the transistor survives (assuming adequate heatsinking).
High power high frequency Class D will need very fast switchers.
By the way, most RF power amps will naturally have a lowish output impedance but impedance transformation at the output brings this up to work OK into the standard 50 ohms. You could ask for a special to be made which can drive a smaller impedance. In any case, full power tuning is asking for problems. You should use an impedance bridge to tune the load, then connect to the power amplifier.
High power high frequency Class D will need very fast switchers.
By the way, most RF power amps will naturally have a lowish output impedance but impedance transformation at the output brings this up to work OK into the standard 50 ohms. You could ask for a special to be made which can drive a smaller impedance. In any case, full power tuning is asking for problems. You should use an impedance bridge to tune the load, then connect to the power amplifier.
Thanks for the explanation.
It is not easy to reduce power during tuning, as the resonance can be very hard to find, and can rise very fast (at high frequencies). Problems occur by mistake; the impedance bridge can be forgotten too...
BTW, what about the 'NXP's Unbreakable LDMOS' (BLF578XR)? Could it be a solution, here?
It is not easy to reduce power during tuning, as the resonance can be very hard to find, and can rise very fast (at high frequencies). Problems occur by mistake; the impedance bridge can be forgotten too...
BTW, what about the 'NXP's Unbreakable LDMOS' (BLF578XR)? Could it be a solution, here?
Seriously, switch in a 10dB 50 ohm pad rated for the appropriate power, tune to resonance then shut down and switch the pad out, this is time honoured methodology for this sort of thing.
Do NOT be believing NXPs claims, the things are tough sure, but you will note that the spectacular video demo has the fault close to the amplifier (So effectively mostly resistive), with the device well cooled and (I suspect, a fast and carefully set current limit). You would still want to trip the thing off line plenty fast at 2:1 or so in reality.
The other approach is to forgo the sand and use glass vacuum fets (Actually usually ceramic) instead, much more robust against short term overloads, and the Pi network at the output can usually match a fairly wide range of impedance. The downside is the high voltages, and need to tune the amplifier tank anytime you move frequency or the load changes impedance, and the noise from the blowers.
A trick for finding the resonance, build a current and voltage transformer (Like a 'tandem match' style VSWR meter, but with separate outputs) and hook a vector voltmeter across them, resonance occurs when the phase shift is zero.
Regards, Dan.
Do NOT be believing NXPs claims, the things are tough sure, but you will note that the spectacular video demo has the fault close to the amplifier (So effectively mostly resistive), with the device well cooled and (I suspect, a fast and carefully set current limit). You would still want to trip the thing off line plenty fast at 2:1 or so in reality.
The other approach is to forgo the sand and use glass vacuum fets (Actually usually ceramic) instead, much more robust against short term overloads, and the Pi network at the output can usually match a fairly wide range of impedance. The downside is the high voltages, and need to tune the amplifier tank anytime you move frequency or the load changes impedance, and the noise from the blowers.
A trick for finding the resonance, build a current and voltage transformer (Like a 'tandem match' style VSWR meter, but with separate outputs) and hook a vector voltmeter across them, resonance occurs when the phase shift is zero.
Regards, Dan.
This is more of a amateur Ham radio operators question.
One e-mail group that deals with low radio frequencies.
Topband Archives
One e-mail group that deals with low radio frequencies.
Topband Archives
I would say that for your need you may consider a LLC switch-mode power converter.
Theoretically you could use a class-d-amp as a driver, but with 1kW upto 1MHz it is preferrable to use a LLC-controller-chip like NCP1396 and a pair of power-MOSFETs to reach your goal.
Tune its operating frequency according to your series resonant tank. And you are done.
Theoretically you could use a class-d-amp as a driver, but with 1kW upto 1MHz it is preferrable to use a LLC-controller-chip like NCP1396 and a pair of power-MOSFETs to reach your goal.
Tune its operating frequency according to your series resonant tank. And you are done.
Any efficient amplifier can be blown-up by large reflected power from a resonant load.
An alternative IS, as Dan says, to put 90% of your power in a big resistor, tapped so maybe 10% of power goes to a happy load, and no huge hunks of reflected power can pass back to the amplifier. For your system, use a 600W 50r resistor in series with a 25W 2r resistor, your L-C tapped across the 2r. A "50r" amp will not mind a 52r load, nor a 50r load. Reflected power must first drive the 2r tap resistor, then back-feed the 50r resistor, before it can harm the amplifier.
A key point is how your load tuning varies with power. An ideal L-C resonates the same at any power level. Radio receivers are expected to hold a station even when it fades in and out. However a coil with BIG power is liable to heat, change dimension, and change frequency.
Too-sharp tuning can be spoiled by reducing the tank "Q". Added series resistance in coil will do that. Here it could even be as "easy" as winding the same coil with smaller wire, more self-resistance, a basic engineering optimization.
That amplifier has power control. It seems to me that if you limit (AGC?) the forward power to 150W, the reflected power can not exceed the reflected power rating.
> Class D amplifiers
The early "audio" class D amps found use for sub-woofers, because their midrange performance was not good. Decades later the midrange is good, but I do not know if they are ready for 1MHz with clean wave-shape. I had thought the switching frequency was near 1MHz(?) so the maximum signal frequency would be significantly less. (2:1 in theory, 4:1 to 10:1 in practice.)
An alternative IS, as Dan says, to put 90% of your power in a big resistor, tapped so maybe 10% of power goes to a happy load, and no huge hunks of reflected power can pass back to the amplifier. For your system, use a 600W 50r resistor in series with a 25W 2r resistor, your L-C tapped across the 2r. A "50r" amp will not mind a 52r load, nor a 50r load. Reflected power must first drive the 2r tap resistor, then back-feed the 50r resistor, before it can harm the amplifier.
A key point is how your load tuning varies with power. An ideal L-C resonates the same at any power level. Radio receivers are expected to hold a station even when it fades in and out. However a coil with BIG power is liable to heat, change dimension, and change frequency.
Too-sharp tuning can be spoiled by reducing the tank "Q". Added series resistance in coil will do that. Here it could even be as "easy" as winding the same coil with smaller wire, more self-resistance, a basic engineering optimization.
That amplifier has power control. It seems to me that if you limit (AGC?) the forward power to 150W, the reflected power can not exceed the reflected power rating.
> Class D amplifiers
The early "audio" class D amps found use for sub-woofers, because their midrange performance was not good. Decades later the midrange is good, but I do not know if they are ready for 1MHz with clean wave-shape. I had thought the switching frequency was near 1MHz(?) so the maximum signal frequency would be significantly less. (2:1 in theory, 4:1 to 10:1 in practice.)
Indeed it is. In a somewhat related 3MHz lab project, we used a tube ham radio transmitter
and a high power antenna tuner to match the transmitting antenna load. This equipment is
relatively low cost, and much faster to get into operation, compared to designing, building,
and debugging your own.
Last edited:
Harmonics would be quite high with the LLC inverter approach unlike an wideband RF "industrial grade" amplifier with inherent low harmonic distortion across the band - that's partly why the cost of a 10 KW 1 Mhz amplifiers are in the $10000 and above range.........
I was just clarifying that it takes some considered engineering effort beyond just whipping up an circuit from an LLC regulator chip app note to produce lo distortion (harmonics below -80dB) 1 KW level power over a continuous (ie without passive filter banks) continuous range of 100 Khz to 1 Mhz - and that is why most people turn to a ready made industrial grade RF amplifier like the original poster mentioned (AG1010).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.