Key difference between class D & E

[W4NEQ]

I tried the Class E forum, without results,  so I'll ask here to a bigger group.

What is the KEY difference between class D & E as it relates to a 80 or 160 meter Fet-based transmitter?  Both typically use square wave or saturated gate drive.  Both seem to resonate the output network.

The apparent differences in the topology are that E seems to be resonating on the primary side of the output transformer, and, the output network includes a series resonant element to present a higher impedance to harmonics.  But I'm still not clear if these are essential to one class.

So,  what is the KEY distinction between the two?

[steve_qix]

I did answer in the class E forum    If it will help, I will put that answer here too.  You may not have seen it.

Regards,

Steve

[W4NEQ]

I read that, - the part about being the same through driving the gate, and the potential differences in efficiency with frequency.  But as far as circuit topology, What distinguishes the two classes?  Specifically:

1.  Is it the output network providing higher impedance to harmonics?

2.  Does Class E _require_ resonating the drain side of the output transformer?

Thanks,
Chris

[k4kyv]

It's been a while since I have done any study of class E, and the few details I managed to grasp have become fuzzy, but as I recall, the class E circuit is somehow configured so that the instant when the square wave pulse flips polarity, is synchronised precisely with the point in time where the RF sine wave in the resonant tank circuit crosses over the base line (Vrf=0). Therefore, the final is not hot-switching rf power in the course of the rf cycle. With class D, these two events are not held precisely in synchronism, so there may be a lead or lag between flipping the square wave and rf crossover in the tank circuit. Class D would be more like class C, except that the output device is supplying a square wave to the tank instead of kicking the tuned circuit with a short pulse every cycle.

Please correct me or clarify, if I am wrong.

[W4NEQ]

Thanks Don,

Fuzzy is a great way to describe my understanding. So does that then define circuit topology?   Which of the following is required for class E and do these two distinguish it from D ?

1.  Is it the output network providing higher impedance to harmonics?

2.  Does Class E _require_ resonating the drain side of the output transformer?

Thanks,
Chris

[steve_qix]

I read that, - the part about being the same through driving the gate, and the potential differences in efficiency with frequency.  But as far as circuit topology, What distinguishes the two classes?  Specifically:

1.  Is it the output network providing higher impedance to harmonics?

2.  Does Class E _require_ resonating the drain side of the output transformer?

Thanks,
Chris

With class E, the output network is resonant and on the inductive side of resonance. To any harmonics, it will also represent a very high impedance.  However, the waveform, particularly in balanced topologies is not overly rich in harmonics.  There is a big capacitor from the drain to ground which, if the MOSFETs were switched on when voltage existed across this capacitor, would cause losses in the MOSFETs themselves.  However, all of the output components are part of a somewhat inductive, tuned resonant circuit, so there are no fast rise and fall times.  The MOSFET is turned on only when the drain voltage has fallen to 0 volts, so there are no switching losses.

This is the distinct difference between class E and class D.  With class D, the output is close to a square wave, and when the MOSFET is turned on, there is (in theory) voltage across the drain and associated capacitances.  When the MOSFET is turned on, there is some loss within the MOSFET due to the current required to discharge the output and other stray capacitance.  The output transformer, in theory, is transparent and does not factor into the resonant equation.  In practice, there is some leakage inductance so this adds to the total circuit inductance.  The output transformer is not required for class E; it is typically used for convenience in coupling to the output network, and for implementing systems with multiple modules.  You could use an RF choke, and couple directly in to the output network and everything would still work the same.

Now, class D works very well at lower frequencies and with square wave drive where these losses are not significant.  As the frequency is increases, the switching losses become greater.  There is a point where the switching losses become considerable, and another topology (such as class E) is necessary to maintain efficiency.

More later!  Regards,  Steve

[WA1GFZ]

W1VD web site has some good information on class D.
I ran both and found class d needs a good resistive load or you get a lot of ringing of the drain signal. Jay has a good handle on snubber design  and drain loading to handle drain ringing. I found the best tuned circuit for class D was a series tuned L/C on the 50 ohm side of the output transformer, but Jay is running low pass filters so no tuning is required. You can get away without tuning if you have a good 50 ohm load.
The shunt drain C in a class E final eliminates all the ringing. The tank effect  swings the drain voltage near zero at the time the gate is turned on as Steve pointed out. The shunt C integrates all the crud and sets up a nice flywheel effect to set up a close to zero volt switch. This helps efficiency and eliminates all the high frequency voltage transients protecting the switching FET. Jay does the same thing with snubbers and a bit of tuning across the transformer primary.

[W4NEQ]

OK, thanks guys,  I'm slowly getting there ...

So, if I understand - the drain side of the transformer is set up with shunt C _right there_ to help dampen ringing (and presumably parasitics),

The secondary side of the output transformer load impedance must be kept somewhere inductive at fundamental to provide a conjugate with the drain C,

Presumably, tuning this L/C controls phasing to achieve the zero voltage switching,

And a series-tuned output provides provides a higher impedance to harmonics to allow switching to be closer to square wave action. 

All of this is independent of subsequent low-pass filtering. 

And you cannot relocate the shunt C to the output side of the transformer (perhaps a step up to 50 ohms transformer) because it would not be effective suppressing ringing and parasitics?

Chris

[steve_qix]

Hi Chris,

Yes, that is pretty much all there is.  You don't need any further low pass filtering or harmonic reduction with class E.  The tuned circuit does all of that for you.

The shunt C, as you pointed out, MUST be connected directly to the drains - or at least some of it must be, otherwise there would be severe ringing caused by the leakage inductance of the transformer and intervening wiring.  In fact, that is the only real low impedance path in the output - the shunt capacitor from drains to ground.

The shut C is really there to integrate the drain waveform over time.  You can, in fact, operate without a shunt C (or have a very small one), however, the drain waveform becomes very "peaky", and the peak to average ratio is quite high, and the drain voltage rises to excessive levels.  In theory, you can run about a 3.5 to 1 ratio - peak drain voltage to DC voltage and not compromise the efficiency.  The peak to DC ratio is almost completely determined by the shunt C, and its value with respect to the RF amplifier impedance (dc voltage / dc current).

The matching function must take place somewhere, regardless of which class you're using.  If you're using class D, you must provide a 50 ohm load (assuming your RF amplifier is designed for 50 ohms), and usually this is accomplished using an antenna tuner.

With class E, the matching function is accomplished in the RF output network, and class E amplifiers can work into a fairly wide range of loads, including a certain amount of reactance, without difficulty.

But, either way there are tuned circuits somewhere.

[WA1GFZ]

Nate Sokal who held the class E patent had a couple great articles in QEX a bunch of years ago. I want to say around 2000. He had an excellent picture of how circuit values effect the drain waveform.

[DMOD]

Here is some info on Class D and E amplifier topologies:

 
Phil - AC0OB

[WA1GFZ]

Look at figure 5 in the Sokal article. It is a great tuning aide while watching the drain waveform on a scope. 

[W4NEQ]

Thanks you all,

Those two articles were very helpful.  Class E sure is an odd beast.

Chris

[steve_qix]

Thanks you all,

Those two articles were very helpful.  Class E sure is an odd beast.

Chris

It's not as complex as the articles might suggest.  Doing any sort of analysis on RF circuits will get complex.  I remember in college, I took a course on designing directional antenna arrays.  I filled *pages* with differential equations and calculus figuring these out to the Nth degree.

Class E is simply a switched amplifier looking into a somewhat inductive tuned circuit.  If you do an analysis of class C it would be similarly complicated    Or try analyzing an analog amplifier with negative feedback - yet we blithely build and use these on a daily basis and think nothing of it.

Regards,

Steve

[DMOD]

I remember in college, I took a course on designing directional antenna arrays.  I filled *pages* with differential equations and calculus figuring these out to the Nth degree.

We did all the E(theta) stuff on computers and calculators.

And I thought you were much younger than me.  

BTW, in our company we are designing and building 4kW ClassE modules for HF and VHF use. The one 4kW cube is
about 15X15X15". That includes power conditioning and tuning.

But I still love the glow of tubes and ionized air.  

Still working at home on constructing one of Steve's ClassE design's. I have all the parts but greatly procrastinate when it comes to drilling holes and punching chassis.  
 

Phil

Addendum: There are a number of good texts out there on RF Power Amplifiers:

Switchmode RF Power Amplifiers by Andrei Grebennikov, Nathan O Sokal

and

RF Power Amplifiers by Marian Kazimierczuk

and

RF Power Amplifiers by Mihai Albulet

and the old standby:

Solid State Radio Engineering by Herbert L Krauss, Frederick H Raab, Charles W Bostian

[steve_qix]

And I thought you were much younger than me.  

It's really amazing what "Just for Men" (that stuff is GREAT!) (and of course getting a lot of exercise) can do to fool people   

[WD5JKO]

Hi guys,

   I wonder if anyone has info (schematics, manual) for the Advanced Energy RF amps such as model 3000, 3001, or 3002? These run the RF stage class E from a 0-200V offline switcher supply built in. They are set for 13.56 Mhz for industrial needs, and are rated at 3000 watts RF out continuous. The efficiency from 208VAC 3 phase to RF out is about 60% overall. Not too bad after three conversions, AC to DC, DC to DC, and DC to RF.

Here is a link showing an AE 3000 in the box:

http://www.ptbsales.com/vacuum/source/advanced%20energy/rfg3000.pdf

Might make one hell of a 20M AM rig. 

Jim
WD5JKO

[W4NEQ]

It's not as complex as the articles might suggest.  Doing any sort of analysis on RF circuits will get complex.  I remember in college, I took a course on designing directional antenna arrays.  I filled *pages* with differential equations and calculus figuring these out to the Nth degree.

Class E is simply a switched amplifier looking into a somewhat inductive tuned circuit.  If you do an analysis of class C it would be similarly complicated    Or try analyzing an analog amplifier with negative feedback - yet we blithely build and use these on a daily basis and think nothing of it.

Regards,  Steve 

Thanks, Steve,

Not so much the math - he's worked it out into cookbook formulas, but how critical the timing appears to be.  Also, he talks about increased efficiency when switching at a 20% voltage minimum - but that stresses the integration capacitor more.

Looks like a good idea to bring out some voltage-divided bnc ports for looking at the drains with a scope.   The BE guy's push-pull arrangement floats the transformer CT with a choke?  I gather that due to swamping the transistor output with shunt C, the rest of the output network becomes less critical as far as preventing parasitics ... a class-E advantage?

For my 375 watt 80 / 160 project I'm going down the copper heat spreader / kapton / heatsink road for my primary low impedance shunt C, but it's not easily adjustable.  I also don't know what to expect in the area of temperature stability.  Still haven't found a source for appropriate shunt padder capacitors - I've seen some very expensive hi-q surface mount types that look low impedance, but are so microscopic I can't see how they could last with much current.   If anybody has a source for the ATC 100 types, like a bunch of 470 pf ...   I'm looking.

Chris


 

[WA1GFZ]

Check QEX Nov 2005. I think you can still download the article free. I've been doing the heat spreader push pull since '95. I did add 1000 pf in parallel with each spreader to get more power out in both the 160m and 75 meter finals.
I have not tried current mode feed to the transformer yet. I think the primary inductance of the transformer may do the same thing. Steve has a good source for RF caps so just undersize the spreader a bit and add caps to dial in the design for the power you want to run. I used door knobs with short heavy leads.

[steve_qix]

I didn't read everything you read, but I can tell you absolutely, having designed and built numerous class E rigs - there is nothing in them that's overly critical.  They are *easy* to tune up and adjust, and the tank construction is a no-brainer.  I build messy stuff with clip leads, and it all works perfectly with class E.

I move frequency all over the place, and change bands almost daily.  It takes me at most 10 *seconds* to tune up the rig when I change bands.  There is only one turned circuit in the transmitter, and that's the output network.  The drive is digital and therefore broad band from DC to 4mHz

The shunt capacitor is fairly non-critical, as long as the capacitor is designed to handle the RF current and you don't introduce lots of inductance in series with the capacitor itself.  Multilayer ceramic caps work well because they are small, and handle a lot of RF current.  I like the ATC capacitors because of their high current handling capabilities (10A for the 100C series), and they can be purchased with a microstrip termination, and they are not expensive.  The termination makes it very easy to solder them into the circuit.

I have never used the the heat spreader / kapton approach because it is mechanically complicated (for me), and being very mechanically challenged, I found it infinitely simpler to just use a capacitor for the shunt, and sil-pads as insulators underneath the MOSFETs and be done with it.  But, the heat spreader method definitely will work.  I will admit it, I'm pretty lazy when it comes to rig construction (much prefer the design phase), so if there's a construction short cut available, I'll definitely take it 

[W4NEQ]

Steve wrote:

 ... "I like the ATC capacitors because of their high current handling capabilities (10A for the 100C series), and they can be purchased with a microstrip termination, and they are not expensive." ...




I'd love to buy some, but I can't find them anywhere.

 

[WD5JKO]

I'd love to buy some, but I can't find them anywhere.

http://atceramics.com/order-products.aspx

[W4NEQ]

I saw the ATC online ordering, but maybe it's just me ...  I cannot see the 100C series available on the drop-down selections.  Only 100A or 100B  - the smaller sizes.

What am I doing wrong?

[steve_qix]

I think you have to call ATC for the 100C caps.  Also, we need to get the MS (MicroStrip) termination.  We could possibly put in a joint order and get a better price due to the higher quantities.  I've done this in the past.

Rules of thumb:  For 4 MOSFETs on 75 meters running 45 volts @ 4A to 5A per module, you need 1000pF per module.  For 6 MOSFETs on 80 meters running 45 volts @ 6.4A to 7.5A per module you need 1500pF per module.  On 160, double the values.  I typically build rigs to cover 2 bands and switch in the additional shunt capacitance when changing to the lower band.  This method works very well.

[kb3ouk]

ok, i've got a good question that fits in this thread. this post: http://amfone.net/Amforum/index.php?topic=29459.msg232875#msg232875 shows a schematic for a single ended class D amp using an IRFP260 on 75 meters, and says will put out 160 watts at 22v 8A. if the amp is made into a class E amp, which looks to be as simple as changing the output network, will the IRFP260 still make the same amount of power? they have a 200v 46A maximum.