Return to KLUBNL.PL main page

[Top] [All Lists]

LF: Re: Effect of LP-filter om efficiency

To: [email protected]
Subject: LF: Re: Effect of LP-filter om efficiency
From: "Vernall" <[email protected]>
Date: Tue, 8 Jun 2004 21:13:52 +1200
References: <[email protected]>
Reply-to: [email protected]
Sender: [email protected]
Dick PA0SE and others,
An interesting contribution.  My pushpull bipolar QRO PA can run to near squarewave performance, and I have successfully used an inductor input low pass filter for many years.  The low pass filter uses three large potcore inductors and HV polystyrene capacitors.  Texts on Class D and E do mention resonant circuits, and most of the commercial applications are in narrow bands, so tuning is a reasonable way to implement Class D/E for those users.  At LF I believe a hard driven final can simply be followed by a low pass filter and still deliver very high efficiency (certainly a lot better than Class C).  In New Zealand our band is about to be expanded to become 130 - 190 kHz, which is a relatively wide band, and not worth having tuning that needs revisiting for significant frequency change.  A broadband PA with a selection in low pass filtering is my preference.  For use only in the 136 kHz band, only one low pass filter is needed.
My preference for an inductor input filter came from reading (I hope my recollection is right) a Motorola article on stability of HF solid state power amplifiers, and the preference was for an inductor input filter.  I merely scaled the filter values to LF and found they worked well.  Possibly the instability bogey is not applicable to LF but I was aware that inrush current would be different if a capacitor input filter was used.  However, there are polystyrene capacitors from collector to ground so on the PA side of the output transformer there are "shunt capacitors" and these certainly reduce the "high frequency nurgles" as seen on a scope probe. 
I have not tried "snubber" series RC networks but I have seen commerial high power switchmode inverters that use that, and the resistors are surface mount with good power rating.  Some of that 16% "harmonic power" gets absorbed in the interests of controlling spike limits.
Power devices (bipolar of MOSFET) can have slower turn-off than turn-on, and I suspect this has a large impact on transients in a pushpull power amplifier.
Regarding suppression of harmonics, and potential radiation, the efficiency of amateur (electrically small) antennas rises significantly (more than linear) with increasing frequency.  So merely a loading coil or "net antenna Q" is only part of the consideration of radiation of harmonics.  The insertion loss of a low pass filter can be kept fairly low and I think it is best to include it in a transmitter design, especially a hard driven "squarewave" machine.
73, Bob ZL2CA 
----- Original Message -----
Sent: Tuesday, June 08, 2004 8:06 PM
Subject: LF: Effect of LP-filter om efficiency

To All from PA0SE

The following subject may have been discussed on the reflector before but I can't remember it.

Class D and E final amplifiers have high efficiency because they produce square waves.  When the voltage between source and  drain of the FETs is high, current is zero; when current flows voltage is almost zero.
But what happens if the transmitter is followed by a low pass filter with a shunt capacitor at the input, as is often the case?
Every cycle the charge in the capacitor must be reversed and with a current of finite value that takes some time. During that time current flows in the conducting transistor whilst voltage over it is not small. So one can expect dissipation in the transistor to increase and efficiency to decrease. 

Looking at it in the frequency domain one could say that the shunt capacitor causes the square wave to start becoming a sine wave.

If the current into or out of the capacitor is so high that charge reversal takes negligible time, so the square wave is preserved, then output resistance of the transmitter is obviously so small it almost puts a short circuit over the capacitor. In that case the capacitor does nearly nothing and could just as well be deleted... 

The problem does not arise when the LP-filter starts with a series inductor. The sudden cut-off of current would result in  a very high kick-back voltage at the drain of a single-ended amplifier, almost certain killing the transistor.
But in a push pull  amplifier this is prevented by the other transistor that starts to conduct at the same moment.
Provided of course coupling between the two halves of the primary winding is very tight.

So it seems to me that class D and E amplifiers should preferably be followed by a low pass filter with a series inductor at the transmitter side.
Please correct me if my conclusion is not valid.

An interesting question is whether a low pass filter is really necessary when the transmitter feeds the aerial via a series tuning coil.
Harry Grimbergen, PA0LQ, has given this some thought and he says the following (my translation).
Not unusual is an aerial capacitance of about 300 pF and a total loss resistance in coil, aerial and earth of some 40 ohms  At 136 kHz this results in a Q of about 100 for the aerial system. At the third harmonic  this produces an extra attenuation of about 280 times, or 49 dB.  The third harmonic of a square wave is 9.5 dB weaker than the basic frequency. On the other hand radiation resistance increases with frequency squared; 9 dB for d3.
Nevertheless harmonics will be suppressed almost 50 dB.

About 16% of the power in a square wave is in the harmonics and this power is reflected back into the final amplifier.
But it is not converted into heat. The DC input is decreased by this 16% as compared to the situation where the amplifier is loaded by a resistive load of 40 ohms for the above case.  I have been able to show this also using simulation by MICROCAP.

A problem in  The Netherlands would be that the radio inspector does not measure harmonics as field strength but as power in the output of the transmitter (or LP-filter, when present). So selectivity of the aerial system does not help.

Any comments will be very welcome.

73, Dick, PA0SE

<Prev in Thread] Current Thread [Next in Thread>