EMRFD Message Archive 7001
Message Date From Subject 7001 2011-12-10 17:30:35 jasonb1963 Load Angle and Phase Delay in CMCD and Class E Amplifiers I spent some more time on the workbench with my CMCD amplifier earlier today and made some interesting observations I thought I'd share.
First, it looks like I have reinvented the wheel that Glenn Dixon already made a year or more ago (though not as well yet). When I saw his schematic I was surprised that it was so similar to mine. However, he had a couple of refinements that led me to achieving over 90% efficiency on 15M with a pair of BS-170s, and I expect that number will rise further as I am still optimizing several parts of the circuit.
Earlier I mentioned load angle with reference to class E amplifiers, which want a -32.48 degree load in order to perform optimally. However, the class D amplifier, like most RF amplifiers, wants to see a resistive (0 degree) load for optimum efficiency. I had previously assumed this tied in with the phase delay of the filters I had been using on the class E amplifiers I was building, but some further reflection now leads me to believe that there are two separate, but related effects here.
The first effect is the load angle, and it is already established that class E amplifiers do not want a purely resistive load -- they must be matched to a complex load. So you can't just design a filter for them with x ohms of input impedance and 50 ohms of output impedance -- you have to design the filter in the complex domain so that it does not alter the load angle.
Where I became confused was in associating this with the phase delay of the filter. I had observed better efficiency with my class E amplifiers when I used a filter which had a phase delay of 0 at the center frequency. There is probably a tie in to load angle here because such a filter will not alter the load angle unless it also performs an impedance transformation.
While working with the CMCD amplifier, I noticed very high efficiency numbers (in the low 90s) with a 50 ohm load and no filtering. As soon as I put a low pass filter in place, the efficiency dropped from low 90s to low 70s. But the filter I used presented a nearly perfect 50-ohm resistive load at 21MHz when terminated with 50 ohms (I don't have a network analyzer, but this is what the NuHertz filter designer software claimed). I would have expected this with a class E amplifier where the load angle is supposed to be -32.48 degrees, but not on a CMCD amplifier which wants a purely resistive load.
So as a next step, I designed a bandpass filter with a zero degree phase delay at the center frequency and put it on the output of the CMCD amplifier. Efficiency was back up to around 90%, just 1-2% lower than before.
Now here's where I get confused. Why would having a zero degree phase delay at the center frequency improve the efficiency? None of the energy at this frequency will be reflected back to the amplifier since it's in the passband and terminated in 50 ohms. Can anyone enlighten me as to what I'm overlooking?
P.S. I don't know if this is significant or not, but without any filter on the output, the CMCD amplifier I built was running between 60-70% efficiency between 10MHz and 50MHz when I tested it earlier. I suspect it would have gone down to 80M too, but I didn't think to set the signal generator that low at the time. I know class C amplifiers can be broadbanded as well, but this seems quite a bit better than the efficiency numbers I remember reading for broadband class C amplifiers. The output is full of harmonics and would need filtering in order to be airwaves-friendly.
7004 2011-12-11 04:51:21 dixonglennb Re: Load Angle and Phase Delay in CMCD and Class E Amplifiers Hi Jason,
I too went through many filter (and driver, and bias) designs and schemes, and found, in my case at least, that what the components were doing outside the passband affected efficiency a great deal. The current switchmode topology has that inductor feeding the amplifier and the tendency is for the drain voltages to ring at VHF and degrade efficiency.
Concentrating on the passband characteristics of your filters/components without regard to what is happening at the higher frequencies may produce many 'head scratching' moments where results do not seem to fit the model in your head.
I am blessed to have a 500 MHz scope and low capacitance probes and found that a goal of keeping a pure half-sinusoid voltage waveform on the drains got me to the best efficiency. You do have to be careful with your probing technique as the probe will have an effect on the circuit, but in my case a standard 10pF probe with the cat whisker ground (a very short grounding wire right at the tip) worked well enough.
I suspect class E would be similar. Whatever filter you use needs to preserve that class E characteristic drain resonance that dips to near zero volts prior to switching. To me this means careful attention to filter input impedances above the transmit frequency.
7005 2011-12-11 07:21:46 jasonb1963 Re: Load Angle and Phase Delay in CMCD and Class E Amplifiers Hi Glenn,
Thanks for your suggestions. I have a 500MHz oscilloscope as well, but my probes are only rated to 300MHz. I will try the "cat's whisker" on the drains and see if there is any VHF ringing present on them.
And to rule out the possibility that the filter's phase delay has anything to do with the efficiency, I'll put another diplexer