Date: Thu, 3 Nov 2005 20:53:20 -0000 MIME-Version: 1.0 X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2615.200 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200 Subject: LF: Re: low pass filter for power use Content-Type: text/html; charset=windows-1252 X-SA-Exim-Scanned: Yes Sender: owner-rsgb_lf_group@blacksheep.org Precedence: bulk Reply-To: rsgb_lf_group@blacksheep.org X-Listname: rsgb_lf_group X-SA-Exim-Rcpt-To: rs_out_1@blacksheep.org X-SA-Exim-Scanned: No; SAEximRunCond expanded to false X-PN-SpamFiltered: by PlusNet MXCore (v2.00) Content-transfer-encoding: 8bit

Dear Tom, LF Group,

There are 2 main requirements for a TX output filter

a) Attenuate harmonics to acceptable level

b) Provide the right load impedance for the PA - usually a pure resistance at the output frequency.

Usually, the impedance at the input of the filter is made the same as the load, I assume 50R. There are a number of ways of doing this:

1) Design a Butterworth, etc. low-pass filter for a cut-off frequency well above the TX frequency - The cut-off frequency needs to be high because the input impedance of the filter becomes highly reactive near cut-off. Fortunately, push-pull PAs produce very little second harmonic, so the filter can have a fairly high cut-off frequency and still attenuate the 3rd and higher harmonics. The G3YXM design appears to have a Butterworth response with f-3dB at 245kHz - it gives about 30dB attenuation at 411kHz, depending on the PA output impedance. The filter input impedance is nearly ideal at (50.2 - j0.2)R.

2) Design an odd - order Chebyshev or elliptic filter with ripple in the passband, such that the TX frequency coincides with the peak of the last ripple before the cut-off frequency, where the insertion loss is exactly 0dB. This condition ensures that the input impedance of the filter will be exactly 50R at the TX frequency. It will be possible to get higher harmonic attenuation this way, but will require more accurate components to get the desired response, with higher ratings due to the higher reactive currents.

3) Design the filter as a matching network with equal 50R input and output frequencies. For example a symmetrical pi-network with XC = Ro/Q, XL = 2QRo/(Q^2 +1). Two of these put together makes a good filter; for example if Q = 1, XC = 50R, XL = 50R, and you get a filter that is 23.2n/58.1u/46.5n/58.1u/23.2n, and gives about 45dB attenuation of the 3rd harmonic (and about 25dB of the second harmonic). This is the "half wave" filter that has been used a lot with HF transmitters. You can get higher harmonic attenuation with greater Q, with the drawback of higher circulating currents / more critical tolerances. You can get quite adequate results with Q as low as 0.5.

4) Design an output network to achieve particular time-domain waveforms, e.g. the class E designs by G3NYK.

5) errm... that's all I can think of now :-)

The 47nF-----> 54uH----->82nF------54uH----->47nF filter does give higher harmonic attenuation than the G3YXM design, and has an input impedance of around (75-j6)R at 137.5k. But reactance and resistance vary rapidly around this frequency, between 50 and 95R and j0R and -j85R, so small component value, load, or frequency changes could lead to the impedance varying a great deal. This is perhaps why it causes some problems.

With a linear PA that delivers a sinusoidal voltage or current to the load, you mostly only have to worry about getting the right impedance match at the TX frequency - but with non-linear amplifiers like class D, you also really need to think about the waveforms at the PA output - the filter effectively is part of the PA. Ideally, in a current-switching class D design (as most are), the load should be a high Q parallel-tuned tank circuit - this requires a filter with a shunt capacitor at the input, allthough in practice the Q is not very critical, and can be 1 or less with only slight loss in efficiency. Voltage-switching class D circuits (e.g. Decca) require a series-tuned tank. Also, some amplifiers need a load that is inductive or capacitive at high frequencies for stability. For example, the modified audio PAs are often unstable with a capacitive load, so require a filter with a series inductor at the input, so a T network instead of a pi network.

Cheers, Jim Moritz

73 de M0BMU

----- Original Message -----

From: Thomas Eckart

To: rsgb_lf_group@blacksheep.org

Sent: Thursday, November 03, 2005 4:21 PM

Subject: LF: low pass filter for power use

Hi Lf's
in all circuit diagrams I find filters like 12nF---->54uH---->27nF----->54uH----->12nF
here is the 3db point at 250khz.
Has anyone test a filter in this configuration 47nF-----> 54uH----->82nF------54uH----->47nF
here is the 3db point at 147khz.
The 2. filter gives a better suppress of the harmonics!
Wat is the reason that all use the fist one?

best 73' Tom DL4EAU