Message-ID: <001701c44d4e$fe7cdc00$4e0a97d4@oemcomputer>
From: "Johan Bodin" <jbodin@tele2.se>
To: rsgb_lf_group@blacksheep.org
References: <6.1.0.6.2.20040608100555.02801340@POP3.freeler.nl>
Date: Tue, 8 Jun 2004 11:51:50 -0000
MIME-Version: 1.0
Subject: LF: Re: Effect of LP-filter om efficiency (long)
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org

Dick,

you are absolutely right that a "voltage switching" class D amplifier should
not be followed by a lowpass filter with shunt C input as this will result in
huge current spikes.

There are two major kinds of class D, "voltage switching" / "constant voltage"
and "current switching" / "constant current". G3YXM's 1kW transmitter is a
good example of current switching class D design. The DC power enters the
center tap of the output transformer through a large DC feeding choke which
keeps the DC input current fairly constant over the entire RF cycle while the
FETs take the current in an alternating fashion. The current through each FET
is ideally a square wave while the voltage across the (open) FETs is a half
sine period. This kind of amplifier *should* work into a parallel resonant circuit
or at least a shunt-C-input lowpass filter. A common problem with this kind of
amplifier is high voltage spikes that occurs when there is a "drive gap" (when
both FETs are off). Snubber networks comes to mind...

Full bridge (like the Decca) and half bridge (totem-pole) amplifiers are of the
voltage switching kind and they work in "the opposite way". The voltage across
the FETs (or whatever active devices) is a square wave while the current is
sinusoidal. These amplifiers should drive the load through a series resonant
circuit. With enough loaded Q in the series tank, usually 5 to 10, the load current
is an almost perfect sine wave. The harmonics exists only as voltage across
the transistors, only fundamental current is drawn out of the circuit.
A properly designed amplifier of this kind has much less problems with overvoltage
spikes since the maximum transistor voltage is confined within the supply rails.
However, under ideal conditions, the FETs will switch at the zero crossings but if
the load is reactive, or if the series circuit is out of tune, the voltage/current phase
shift will increase switching losses. Also, the transistors will have to conduct
"backwards" during a part of their ON period. MOSFETs may survive this thanks
to their inherent "backward diodes" (slow! lossy!) but bipolars can quickly enter
class Z (flash & bang!) if no external diodes are there to catch the backward current... :-)

Andy G4JNT has written an excellent article about his 700W class D amplifier.
The elegant Decca overcurrent protection is also described in this article.
The article is available in PDF format on this page:
http://www.nutstreet.net/xfx/smt/


Class E is very different from class D. Even though the drive is a "hard square",
neither the transistor voltage nor current is a squarewave. Like class D,
class E approaches 100% efficiency but, as opposed to D, class E requires
quite a big capacitor across the transistor which minimises problems with high
voltage spikes. Tuning is easy, simply adjust the shunt C and the series tank
(either C or L) for a proper drain waveform.

In my opinion, class E is the simplest and most elegant amplifier. The main
drawback is that the "transistor utilisation factor" is less than for class D i.e.
the same transistor would be able to give more power in class D.
In class E, the peak transistor voltage is 3.5 - 4 times supply voltage and
Pout is only 0.55 * Vsupp^2 / RL (approx.).

If the presence of even harmonics is a problem (due to assymetrical drain
waveform), it is possible to make a push-pull class E amplifier by building two
identical single ended class E amplifiers, feed them 180 degree drive, and
let them share a common series resonant tank and output transformer.
Thanks to the series tank, the amplifier halves will never "know" that they
are operating in push-pull since they both "see" a nice sinusoidal load current.

Nathan Sokal WA1HQC has written a very good article about class E design.
The article sokal.zip (zipped PDF) can be found on this page:
http://mywebpages.comcast.net/tonne/classe.html


> 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.

Maybe you could get away with a small class E amplifier. The voltage waveform
on the transistor is not too far from sinusoidal. The series circuit required in
class E can consist of the aerial system itself since there is no upper limit on
loaded Q in the design equations.

It may be tempting to connect the loading coil + antenna directly to the
midpoint of a (voltage switching) class D totem-pole, but, because of
nanosecond switching speed, the wire running from the amplifier to the
loading coil will radiate a lot of harmonics! All the way up to VHF...
(I tried it once and caused a TV blackout :-)


73
Johan SM6LKM