MOS or IGBT class H modulators for tube AM transmitters.

[IN3IEX]

In the past I studied a new kind of parallel class H ( class H// ) power voltage followers for plate and g2 AM modulation and general purpose high voltage applications.
I found a very simple method to combine DC power sources with different voltages and achieve high conversion efficiency.
For AM high voltage modulators, with four supply voltages we have efficiency of a few percent higher than the classical transformer modulation with class B audio power amplifier.
I obtained proper current switching among the various source followers by using a different number of series diodes connected to the sources of the Power MOS (or emitters of IGBTs). See the attached schematic.
The circuit worked in the simulator (SPICE). Using 4 MOS, the one at lower supply voltage will have 1 diode in series to the source, for the nearest (next) higher supply voltage level add a diode, and so on, one more diode per level. MOS turn on voltages must be matched or add two diodes per level.
The MOS with less diodes will conduct first and take all the current, but if it does not have enough supply voltage, the next MOS will provide the current and will switch off the first MOS because of the drain and source diodes, and so on. Very simple.
I think that reasonable voltages are 300, 600, 900, 1200 V that can be obtained from 220V AC with voltage multipliers.
1500V MOS or IGBT are available and the rest voltage could be set (and obtained with a voltage divider) to 95% of the 600V supply (instead of taking it from the 12V) to have low dissipation without modulation and actively track this voltage.
This circuit could be used to modulate 807s, 6146s, sweep tubes,  (1, 2, 4 in parallel) etc. In principle it can provide a lot of current, much more than a common modulation transformer. Higher audio quality is also possible and low level audio clipping/compression will work perfectly because it is a nearly ideal voltage source.

Who will test it?

Giorgio

EDIT: I changed the name to class H// (after Steve's note).
Maybe it is not necessary to match the MOS/IGBT, just select them and use the one with the lowest turn on voltage for the lowest supply voltage and go up.
The class A driver is the simplest driver that may work "safely attached to ground". The gate capacitances are an issue indeed. But keep it extremely simple. With 5-10A 1500 V IGBT this driver is fine. Obviously we must add gate protection circuits (zener diodes...)
Considering the dissipation of the class A driver, the best "energy efficiency"/"gate capacitance" compromise is obtained with three MOS/IGBT. Three supply voltages: 1200, 900, 600V.   300V are removed. I evaluated the dissipation of each of the four MOS with a sine signal with 580 V baseline, and removing the 300V supply was not an issue, but saves driving power.
When dealing with such high voltages, simplicity is important. After a spark, if the diodes do not fail, they will protect the various MOS/IGBT from the failure of a single MOS/IGBT. Failure of one power supply may only partially affect the output and could be detected in time.
The power supply that I designed and made is the last one of these:
http://www.ing.unitn.it/~fontana/Comparing%20Half.pdf  picture attached, it is 10x16cm and provides 300, 600, 900, 1200V. It is now used in another TX.
The modulator that I did not made (till now) had to be 10x16 cm

I have these IGBT:
http://www.ebay.com/itm/IGBT-SGF5N150-1500V-10A-Fast-15ns-Trise-Insulated-x4-/180713394575?pt=LH_DefaultDomain_0&hash=item2a135b918f#ht_2177wt_901
http://www.fairchildsemi.com/ds/SG/SGF5N150UF.pdf
The input capacitance is reduced by the fact that it is an emitter follower. The output and the reverse transfer capacitances are there with their values but the effects are affected by the switching and the capacitances of the diodes.
Fortunately when the output voltage is very high and the current through the load resistor of the class A driver is low, only one MOS/IGBT is connected to the driver because all the other MOS/IGBT are disconnected by the diodes.

[WA1GFZ]

put some 15 volt zeners gate source to protect them. This is similar to a design I have for class E low voltage. You can extend the high frequency response with a small cap across the feedback resistor. I've seen IGBTs rated for 4000 volts.

[steve_qix]

You may have problems with the circuit, as drawn   , but it's a very good idea!

I've been kicking this around for a long time, and even have the 2500V IGBTs here to build it!

Several things - this is not class H      In class true H (and its related cousin, class G), there are two devices in series.  The "lower" device is operated as a source follower, and is close to, but NEVER achieves saturation.  The 2nd (upper) device is not conducting at all at carrier, or during "negative" modulation, and the "lower" device gets its drain voltage through a diode from a power supply, as you have it in your circuit.  

The source (through a protective / equalizing resistor) of the "upper" MOSFET is connected to the drain of the "lower" (linear) MOSFET or transistor.  The gates are not operated at the same DC voltage.  The "upper" MOSFETs gate is set such that a small increase in positive-going energy will turn that device on (and subsequently begin supplying more voltage to the "lower" MOSFET) before the lower device saturates.  So, the lower device is always operating in the linear portion of the curve, and is the main active element.

The significant advantage of doing it this way, is that there is no crossover point and therefore no crossover distortion is possible.  The lower device is ALWAYS operating, and in linear mode, and  [in theory], the output voltage is ONLY affected by the gate to source voltage and NOT by what is happening at the drain. In the real world, the reverse transfer capacitance (the Miller capacitance) will "pull" on the gate at high frequencies as the drain voltage is varied.  This can be mitigated by using a low impedance driver, and providing a driver "bus" which is independent of the output.

In my latest class H designs, I have incorporated this improvement, and have eliminated the effect of the Miller C on the output, at least within the audio range.  Bipolar devices have significantly lower miller C, and are therefore often used in very demanding class H and class G applications.  I have also seen pulse width modulators used to supply the voltage to the linear element - giving the efficiency of PWM and the clean, unfiltered performance of class A - the best of both worlds!

If you use a pull up resistor - class A voltage amplifier model (as you have it here, more or less), it may be desirable to use an additional source follower ahead of the output devices.  This will provide a level of "swamping" of the gate capacitance, which is rather significant in large, high voltage devices.  The R/C time constant of the pull up resistor, with the combined capacitance of the output MOSFETs will fall squarely in the audio range, and this is undesirable for a number of reasons (the most important of which is phase shift).

Another way around the problem, and one which will also remove the need for any pull up resistors at all, is to use a floating op-amp with an independent bus for the floating common, therefore drastically reducing the effect of the Miller C on the output.  This is due to the fact that the driver bus will not be connected to the output and therefore, not subject to the effects of the miller C at all.  This may sound very confusing without a schematic, and, unfortunately I have not yet written this design down (although it is built and running very nicely!).

You use the common mode rejection characteristic of the op-amp to remove the DC component caused by floating the op amp with respect to DC ground, from the output.  This keeps the entire amplifier DC coupled.  You can incorporate an error amplifier and reference at low level, and use a high amount of DC, and a lesser amount of AC feedback, resulting in a rock stable, and low distortion system.

I realize this may sound confusing without a diagram, and I will try to make one up, and post it in a subsequent reply.

Suffice to say, however, that this is a good idea, and one whose time has come.  There are very practical devices available, and such a modulator would not be overly complex.

Also, the same output devices can be used as a source-follower pulse width modulator (just like the ones in the class E rigs, only higher voltage), eliminating most of the heat, simplifying the power supply, and reducing the supply requirements by more than 50%!!  A pseudo-active pull down would put icing on the cake, making for one really great modulator.

Regards,

Steve

[WA1GFZ]

Steve I don't think he will need a source follower since only 2 FETs unlike the class H at low voltage with many FETs in parallel. This will limit high frequency response if it is an issue. The point about having the voltage or circuit above the drain of the bottom FET will make for a cleaner output. Then the upper FET gate just needs to be 10 or 20 volts above the bottom gate so it never starves. Also prevent the upper gate from going below the upper voltage.
Simulation is so much easier than hacking on the bench.

[steve_qix]

Hi Frank,

I suppose it depends on what one chooses for devices, but whatever they are, the MOSFETs (or IGBTs) need to have significant dissipation, so that the SOA is not exceeded during any portion of the audio cycle.  I wrote a program to calculate this, and it is remarkable how much transient power is dissipated !  Anyway, devices such as these have a LOT of input capacitance, unfortunately.

Using two IXLF19N250A devices as an example (2500V, 19 Amperes, 250 watts dissipation), the capacitive reactance of the two gates (appearing in parallel at audio frequencies) at 10kHz will be around 2600 ohms, which is significant when compared to the value of the pull up resistor used. 

I will add that the ratings of a single IXLF19N250A in the top position of a class H modulator will be exceeded when modulating a 250 watt power input vacuum tube operating at 650 volts, and with a 1900V positive peak power supply (allowing for around 200% positive modulation) (say, 3  6146Bs, such as are in a Valiant).  The device will be dissipating 500 watts of power the instant it begins to conduct!  So, at least 2 devices in parallel would be required, and I would actually recommend 3 devices.

From the experience of class H modulators in the class E world, the modulator devices chosen for the class H modulator each have a dissipation rating of 300 watts.  When modulating a 360 watt carrier (45 volts at 8 amperes), the modulator MOSFETs will get hot under heavy modulation, and I use 4 of them in each bank!  That's 1200 watts of available dissipation per bank.  Consider that the transient power dissipation in the upper bank of class H modulator devices is 720 watts (using 45v at 8 amperes as an example, with a 135V positive peak power supply).... so 4 devices is not overkill by any stretch of the imagination.

Anyway, this can all certainly work in the tube world, just as it does in the solid state world.... although when you start looking at dissipation numbers in analog circuitry such as we are discussing, it certainly makes PWM very attractive, with its 95% modulator efficiency!

[WA1GFZ]

So true.
A screen modulator would be lower power.

[IN3IEX]

Steve, I wrote a copyright paper.....
http://www.ing.unitn.it/~fontana/ParallelClassH.pdf
After writing it .... another wonderful idea. Why not use it for a high efficiency HF linear amplifier?
It should work and stay cool.

What do you think?

Giorgio

[steve_qix]

Steve, I wrote a copyright paper.....
http://www.ing.unitn.it/~fontana/ParallelClassH.pdf
After writing it .... another wonderful idea. Why not use it for a high efficiency HF linear amplifier?
It should work and stay cool.

What do you think?

Giorgio

Hi Giorgio,

This idea has been around for a long time (and its an interesting one) - at least since the 1970s - maybe earlier.  It has been discussed for both tube and solid state circuits.  In fact, I can probably find some diagrams from 40 years ago if I look.

These types of arrangements will "work", but what you run into is crossover distortion.  Feedback compensation helps, but when a lot of feedback is applied to otherwise distorted systems, you generate more transient intermodulation distortion.  I don't see any advantage over standard series class H, and there are some disadvantages.

The commercial short wave and standard broadcast transmitter engineers were probably the first to look into anything that would eliminate the modulation transformer, and increase efficiency.  These folks have essentially driven the high efficiency transmitter design efforts, as very large amounts of money can be saved by creating a more efficient transmitter.  The need for better audio is obvious.

In the designs I came up with that operated in the same manner, I called it "Class B Series Modulation"    It is most like class B, in that the devices that deliver the power to the load are operating with an approximate 180 degree conduction angle, and the devices are not all operating at the same time.  This is most akin to class B, as opposed to class H, where one device operates all the time, and its voltage source is varied, depending on the output at the time.   

Regards,

Steve

[IN3IEX]

I am very optimistic. With 1 kV peak output, crossover distortion is 0.1% with no feedback. With 20 dB feedback should be about 0.01%. No problem.
I am confident that it will work well as modulator for a parallel of 807 using the cited IGBT and the suggested schematic with 3 supply voltages.

For the professional market nothing can beat PWM. But that is more expensive and difficult to shield with limited resources.

Here we are... something to test. Will it work?

[steve_qix]

I say, build it and test it 

And, no matter what - never give up!

[WA1GFZ]

it should work if you set up the simulation properly.

[kb3ouk]

is this circuit more efficient than this modulator: http://amfone.net/Amforum/index.php?topic=27856

what practical limits is there with this class H circuit? with the series modulator in the link, the input voltage to the modulator has to be at least twice the voltage to the  final, so to make at least 100% modulation with that circuit, if the tube is running 500 volts at 250 ma, the modulator wants at least 1000 volts at 250 ma. i assume that the class H circuit needs the same power requirements, meaning the voltage to the modulator has to be higher than what is being fed through to the final? it's not clear to me if the full 1200 volts that the power supply is delivering is passed through to the tube final. is it, or is there a voltage drop through the modulator?
shelby

[IN3IEX]

Hi kb3.

Yes, this "possibly new" classH// is much more efficient than the modulator of that link.
What is important is that W1FYO modulator is READY to be only slightly modified to TEST this classH//.
Thank you for this.
Just add some diodes and a half HV supply, then adjust VMOD Level in order to have unmodulated output voltage 50v less than half HV.
At rest with 100mA anode current the W1FYO design dissipates 500V * .1A =50W, the same design modified for classH// will dissipate 50V * .1A = 5W .....ten times less. With modulation the gain in efficiency is the same we get from the usual class H.

If you can, please call W1FYO and show this discussion. He will certainly be interested.
Maybe we can conclude in a couple of weeks... with W1FYO running this thing for the first time.

Giorgio


 

[IN3IEX]

it should work if you set up the simulation properly.

FOR SALE:  USB Smoke generator: include software for most SPICE implementations. Library of smells attaches directly to part library.... used but still new.

[steve_qix]

it should work if you set up the simulation properly.

FOR SALE:  USB Smoke generator: include software for most SPICE implementations. Library of smells attaches directly to part library.... used but still new.

If it can generate the smell of money, I'm interested!  It'd be great for attracting girls    It would most likely work better than my usual line "Hey honey, I've got the biggest ........................................ class E rig you've ever seen!".