Simple approach to matching PA devices of unknown input impedance and output load requirement

There are plenty of surplus high power cellular transmitters available at low cost, and these are usually full of useful components that would cost a fortune if bought new direct from a stockist. Of particular interest here are the RF power output devices. A look at the data sheets for most of these devices will show RF impedance information only at the cellular band(s) frequency for which they are intended. This does not mean that the device cannot be used elsewhere, particularly at lower frequencies, but it does mean that data will not be available. This applies to both unmatched and internally matched devices - the internal matching usually being restricted to minimising port reactances by incorporating the beam lead interconnect into a low pass T section.

2m PA and driver

Whilst devices will work reliably down at HF frequencies, it is operation at VHF and UHF that lends itself ideal for the matching method described here. By good fortune, this method is inherantly kind to the transistor, as will be seen a little later.

Quarterwave coax
              line
First, some basic transmission line theory:

If a line of characteristic impedance Zo is terminated in a purely resistive load that is not equal to Zo, at a frequency where the line is a quarter wave long, the impedance seen at the other end will also be purely resistive - though it's value will have been changed. If the load resistance  is lower than Zo, the input resistance will be higher than Zo by the same amount (and vica verca), as the table opposite shows.  In other words, the transform see-saws about Zo.


Paralleled lines

Paralleled lines

Putting two lines of equal Zo (and length) in parallel, produces a combination that is equivelent to a single line of characteristic impedence Zo/2. This is easy to demonstrate by considering the two lines as seperate circuits, with their own terminations, determining their individual transforms, then paralleling them and again considering the new overall transform.

transform table
So using paralleled 50 ohm lines will provide the transform possibilities shown in the table to the left. Only the transforms reducing 50 ohms downward are tabulated, since for this application and with current devices (usually FETs of one sort or another) at VHF/UHF, the transform required will always be to something less than 50 ohms for peak powers of 5 watts or more.

So far, only pure resistive transforms have been considered, yet mixed reactive/resistive device impedances will be met in reality, and the reactive part will have to be taken into account. This can be done in practice by considering what happens if a line of less than a quarter wave is considered:

shortened lie

With the line now shortened, the frequency at which a purely resistive transform occurs is higher. However, a capacitor at the open end will lower that frequency and a value can be found that again provides a purely resistive transform at the original frequency. A by-product of this end capacitor is that it offers a way of tuning out device reactance, which as was mentioned earlier, will always be present. The shorter the shortened-line is made, the greater the available compensation will be - but are there any compromises that must be made?.

Two that come to mind are:

a) is the resistive transform ratio altered?

b) is the bandwidth affected?

Some ptfe cable with a Zo of 28 ohms was available (a story in its own right) and a resistive potentiomer was soldered across one end and adjusted until 50 ohms was seen at the open end. The pot value was measured and plotted below for various percentage lengths of line relative the quarter wave length. For all of the shortened lines, a trimmer capacitor was used in place of the fixed capacitor previously mentioned and varied until the purely 50 ohm input condition was obtained.

As can be seen, the transform remains reasonably constant.
change in resistive
              transform
bandwith
              variation

For the two extreme cases measured (ie, full quarterwave- wave and 56% of full quarter wave) the return loss (match) was measured to see the effect on bandwidth (see above right). The 20dB return loss points were chosen for these measurements, since this represents a degree of mismatch that can be regarded as negligible. For the shortened line, the bandwidth has reduced by almost a half.


Applying all of this to a 146.5 MHz low distortion power amplifier suitable for datv use:

2m datv amplifier

This is the circuit of the 2m datv amplifier shown in the opening picture. When underun at 20 watts PEP two-tone testing, the third order products are -50 dB relative to either tone (about -30 dB at 70W PEP). It uses a Motorola MRF5S19150 N-channel enhancement mode lateral MOSFET, characterised between 1900 and 2020 MHz and rated at 100W CW from a 28v supply. The amplifier shown runs off a 19v lap-top PSU, hence the lower PEP output at -30 dB intermod performance.

Input parallel lines are added until the match looks good. Bias is added via a resistor, to avoid any unforseen spurious resonance situations that might occur with a feed choke, and which may cause amplifier instability. Amplifier gain is high (about 18 dB), so it is well worth making this resistor a low value, anywhere from 22 to 100 ohms, and accepting a small amount of gain loss.

Adding paralleled lines on the output section is also undertaken, monitoring both the supply current and output power for each addition. Initially, supply rail clipping will result in a relatively low supply rail current being drawn, even with over drive on the input. This is because the 50 ohm load is only being transformed down a small amount. As lines are added, the output power will be seen to increase, as will the supply current. Thus, the approach is inherantly safe, or device friendly. The same logic also explains the use of a full quarter wave
section to feed the supply rail to the device drain, since it will look like a high RF resistance at the frequency of use. With this arrangement, stability seems to be excellent. Likewise, a similar amplifier at 70cm seems to operate in a stable manner, and is currently in use as the (25W) PA in GB3FNY, this time running in Class C.

Finally, a comment about the amplifier shown in the picture. This used the unknown source of ptfe cable refered to earlier that actually had a Zo of 28 ohms, so when using 50 ohm cable, more paralleled sections would be required to get to the same situation. Eigth inch-ptfe cable is currently available from China vie ebay quite cheaply, and being ptfe insulated is very forgiving of lots of soldering.

two tone measurement

--- oOo ---