Re: Excellent presentation on Class E Amplifiers and why the QCX finals sometimes can get fried


Hans Summers
 

Hi all

I am not an electronics engineer. All is self-taught. I studied physics and my former career was 22 years in bank IT (software). I know there are some "proper" engineers here, I am just a guy who loves electronics and radio and muddles along as best I can, and somehow tries to get somewhere and after a lot of blood, sweat and tears, eventually manages something at least. 

Now for Class-E... over the years I have made several attempts at it. I studied the maths. I tried the spreadsheet models. I tried building published designs. I never was successful. I just could never measure the high efficiency Class-E is supposed to give, and of course that was also verified by the temperature of my transistors. 

When developing QCX I decided for the sake of low cost (no need for a heatsink, and no need for big beefy power transistors) I had better try one more time. I developed what I now call "ghetto" Class-E because it is such a simplified method of choosing the components... yet, it does seem to work very well... 

This is a short extract from some text I wrote on the topic of Class-E for part of my presentation next month at Dayton FDIM:

"Some excellent background reading are two papers by Paul Harden NA5N:


Paul NA5N describes two defining features of Class-E:

1) Use of a square-wave drive to reduce switching losses: the transistors are either on, or off… no lossy region in between
2) Reducing the effects of the transistor capacitances. Class-E has a resonant tuned circuit. The capacitance of the transistors, normally an unpleasant lossy aspect, is now a part of the tuned circuit. 

Class-E also has a reputation for being difficult to achieve. All those intense mathematics Google will help you find, don’t help. In reality, once you realise the secret – it is not so difficult. My “ghetto” design process for a Class-E amplifier is simple. Perhaps it is not totally optimal and a few more percentage points of theoretical efficiency could be squeezed out by the more advanced mathematical treatment. But for the average ham, my method produces excellent results with a minimum of fuss! I had previously attempted more complex methods and had always failed. 

Calculation of the impedance of a resonant circuit is simple, and there are many online calculators which will do the job for you. For example, http://toroids.info/T50-2.php which allows you to type in the operating frequency, and the desired resonant circuit impedance. Then the calculator computes the required inductance, capacitance, and the number of turns required for a certain toroid (in the QCX, a T37-2 is used). 

First choose the output impedance. We usually choose 50-ohms, because this is the input impedance of the Low Pass Filter we will use. The online calculator will tell you what inductance is needed, and how many turns to wind on the toroid. The online calculator also tells you the required capacitance to bring it to resonance at the operating frequency. Here we resort to experiment, because it is a little difficult to know what the output capacitance of the transistor is. The device capacitance varies depending on supply voltage and whether it is on or off. A simple experiment is required, adding different small capacitances to the circuit, and measuring the efficiency (measure supply voltage and supply current to calculate power input; then measure RF power output. Divide one by the other to get the efficiency). It is easy to find what additional capacitance is required to peak the efficiency. The resonance is quite broad and non-critical. "

The component values also come out close to some examples Paul NA5N has in his documents; so that's comforting and you feel you are on the right track :-) 

73 Hans G0UPL





On Sat, Apr 28, 2018 at 4:42 PM, Glen Leinweber <leinwebe@...> wrote:
Alan,
A key difference from the classic Class C PA circuit is the lower value of inductance that feeds DC power to the MOSfet(s). In the classic circuit, this inductor is a choke, having high impedance at the operating frequency. For Class-E, it is a much lower impedance, so that it stores significant energy while the MOSfet(s) are ON (for half a cycle or less). When the MOSfets switch off, this energy is available for output power.

The word "flyback" comes to mind. During the half-cycle-or-less that the MOSfets are ON, current rises continuously in this inductor. After switching off, this inductor current dumps out in the form of voltage: MOSfet drain voltage swings far above the +12V DC supply voltage.

The LC low-pass filter feeding the antenna is quite conventional. Because the MOSfet drain voltage is somewhat pulse-like, harmonics are large - the filter must work hard to reject these harmonics effectively. This can be done various ways: higher Q, or an extra stage.


Join QRPLabs@groups.io to automatically receive all group messages.