Link Coupled Output Impedance

[k4kyv]

LINK COUPLING IN TRANSMITTERS

When the same coupling link is used to cover several bands, as is usually the case with swinging link coil assemblies such as the HDVL series, normally the same link is used for all the bands operated by the transmitter. A possible exception would be special versions that include a plug-in link on the coil assembly. Those are seen occasionally at hamfests, and would not be too difficult to build.  But swapping out the link coil along with the main coil makes band changing a real PITA.  A better solution is used with the BC-610: each coil has its own built-in link, allowing the link to automatically change along with the main coil.  Coils with fixed links are no problem since the link is normally inseparable from the coil.


MY LINK-COUPLED TRANSMITTERS

I never found using the same link on several bands to be a problem.  Years ago,  when I  ran a dipole with open wire tuned feeders strung between two trees near the house, I used plug-in coils in the balanced tuner for 80, 40 and 20. A different antenna was used for 160. The main transmitter (the HF-300 rig) has a 4-turn link at the output that is used for all bands.  I simply made a similar diameter 4-turn fixed link for each of the antenna tuner coils, and it worked beautifully on all bands with plenty of coupling, using series or parallel tuning at the ATU as appropriate.

I still use that same 4-turn swinging link in that transmitter, and the smaller 8005/805 rig also uses a 4-turn swinging link, although the plug-in coil tuner in the shack is no longer used. I came up with a different tuning arrangement for the present antenna system, that works on all bands with the link in each transmitter.


BUILDING A COUPLER THAT WORKS TO A FLAT (UNTUNED) PARALLEL-CONDUCTOR OPEN WIRE LINE

Presently I am working on a matching network to couple the link outputs to the new 450-ohm (actually, Zo= 438 ohms) homebrew open wire line (#8 copperweld 2½ inches apart). For test purposes at this point, the load is a 450Ω dummy load rated for at least a kw of carrier at 100% duty cycle. So far, tests have been run only on 160m. The test transmitter is the 8005/805 rig, normally used for 40m.

I tried a link-coupled tuner with fixed link wound tightly over the ATU coil, and tapping the load down on the coil, but couldn't get close enough coupling to fully load the transmitter to rated plate current no matter where the taps were placed. Then I tried replacing the link by tapping both the transmitter output and the load across a number of turns on the coil, moving both sets of taps around the coil in trial-and-error.  I could get good coupling, but if the rf ammeter is anywhere close to accurate, the efficiency appeared to be very poor, and the setting of the tuning capacitor (a 900pf/900pf split stator cap) very sharp.  Next, I tried a balanced L-network, using a split coil with only about 9 turns total, and 825 pf of fixed capacitance in parallel with the variable (both sections of the split-stator variable are in series for a balanced circuit, maximum 450 pf).  That gives the best result, with good coupling and good efficiency. At about 450 watts DC input, the rf ammeter now reads 0.89 amps into the 450Ω load, as opposed to 0.7 amps with the previous arrangement. I suspect the rf ammeter reads a little high, due to not being mounted on a steel panel.

A nearly identical circuit to my balanced L-network can be seen in a mid 1950s vintage ARRL Handbook in the chapter on transmission lines, under the heading "coupling the transmitter to the line". The diagram is shown under the title "alternative circuit for small links". One difference: in the Handbook article they show a chart for  the amount of capacitance required in the tuned circuit for a Q of 2.  For 450Ω on 160m, they recommend 450 pf total capacitance across the coil.  With that capacitance, I could not get enough coupling, so I experimented with placing fixed mica transmitting caps across the tuned circuit.  With the extra 825 pf, I can get good loading, and by adjusting turns on the L-network coil, I get coverage of the entire band, 1.8-2.0 mc/s plus a little margin at each end, using the variable cap. That means the tuning capacitance varies between about 860 pf and 1275 pf, as opposed to the 450 pf shown in the Handbook chart.


IMPEDANCE AT THE COUPLING LINK OUTPUT TERMINALS

I decided to measure the output impedance right at the swinging link.  Another rf ammeter was inserted in the line between the link and one side of the L-network.  With 0.89 amps going into the dummy load, I get 4.5 amps at the link. Using the formula P=I²Z, assuming 10% loss in the L-network, and plugging in the known figures on each side of the equation (dummy load impedance, rf current to dummy load and rf current at the link), I get a little over 18 ohms impedance at the link. Calculating with zero losses in the L-network gives around 17 ohms.

This is not unexpected, since the coupling link has only 4 turns and the frequency is 1.8-2.0 mc/s.  This also explains why only 450 pf resonating capacitance will not give high enough Q to achieve full coupling to the load. At this frequency, the Q is  probably much less than 2 (I have not attempted to calculate it).  When the same link is used over several adjacent bands, the output impedance can be expected to increase substantially from one band to another, as you go higher in frequency.


USING LINK COUPLING TO A FIXED, PURELY RESISTIVE LOAD

It would be very difficult to get the same link to work over several bands, working directly into a fixed load such as 50Ω coax, without a variable capacitor to tune the link to resonance. At 160m, a small link probably will not work at all using only a simple variable capacitor in series.  I had to use a capacitive voltage divider with mine, using a couple of .003 mfd fixed micas in series, trimmed by a variable capacitor for exact resonance. The series variable cap works fine on 75 and 40.

The next phase of the project is to repeat the above experiments for 80m and 40m.  Detailed notes are being taken, and the final band-switchable version of the tuner will be built, based on the data.

[K5UJ]

Don thanks for writing up your latest work; it is a lot easier to read it so I can digest it at my slow pace than try to get an understanding in conversation.  I am looking forward to hearing how things play out but for now, it seems band switching will be a challenge -- I'd probably wind up with the easy but very costly way out -- 2 or 3 separate networks switched with motor driven RF contactors.

[KC2ZFA]

Thanks for sharing Don. In your opinion, with a variable link and a capacitor to ground does one really need a tuner when feeding 50 Ω coax (to a dipole, vertical, or beam) ?

Peter

[k4kyv]

The variable link with a series capacitor to ground should be enough; the additional tuned circuit in a link-coupled tuner will just waste power.  The capacitor tunes the link to resonance.  That's all the tuner you need.

But that wouldn't work with either one of my rigs on 160.  4 turns on the link isn't enough for 160.  I could get it to resonate, but could load the transmitter only to about 1/3 plate current.  Using trial-and-error, I ended up with a couple of .003 mfd transmitting mica caps in series, strapped across the link.  One side of the link was grounded, and the inner conductor of the coax went to the point where the two series caps are tied together, so that the 50Ω line was connected across one of the mica caps. A 1800 pf variable cap was connected across one of the caps to act as a trimmer to peak the whole thing to resonance.

To tune the transmitter, first run the link all the way out, and with the transmitter in tune mode, dip the final. Then run the link in slightly, and tune the capacitor until you get a slight peak. Then run the link in until the final loads up to full plate current. Re-check the dip, then switch to full transmit mode, and touch up the tuning if necessary for proper plate current.

[KC2ZFA]

that's great info. So, if I understand this correctly, on 160 with, say, TVL or HDVL coils one would need a 5 or 6 (or more)  turns link if one wants to stick with the series cap (of enough capacitance) to ground on the variable link. But in that case there's a spacing problem (not enough space between the two halves of the ready-made PA coils) so one should go homebrew a coil/link for that band. Right ?

[aa5wg]

Don:
Thank you for the information on your work.
Chuck

[k4kyv]

that's great info. So, if I understand this correctly, on 160 with, say, TVL or HDVL coils one would need a 5 or 6 (or more)  turns link if one wants to stick with the series cap (of enough capacitance) to ground on the variable link. But in that case there's a spacing problem (not enough space between the two halves of the ready-made PA coils) so one should go homebrew a coil/link for that band. Right ?

I have one of those plug-in links for the 500-watt swinging link version of the coils (don't remember  the type designation, but a smaller version of the HDVL series.  It must have a dozen turns, wound in two layers.

The problem with the ready-made versions is there isn't enough room in the gap between the two halves for more than about 4 turns of round wire of sufficient gauge, and still have a safe amount of spacing between link and coil.  In my  transmitter, the fully modulated +HV is on the main coil, so there has to be a certain amount of spacing to avoid arc-over. One way of getting in more turns would be to spiral-wind the link, rather than the usual helical (solenoid) winding. Another possibility would be to used an edge-wound coil for the link, since more turns can be made to fit into a smaller space but still have sufficient surface area on the conductor.

Usually, the ready-made coils came from the factory with too many turns on the main coil, since they were designed to accommodate the greatest likely voltage-to-current ratio for the final, and it was expected that a careful builder would remove a few turns until the desired L-C ratio was achieved. Usually those turns were removed from the outer edges, but if a turn or two is removed from the inner edges of each half of the coil, more space is left available for the link coupling coil.  You don't want to make the gap too large, so that you lose too much of the magnetic coupling between each half of the coil.

A transmitter is described in the 1934 ARRL handbook that shows a wide gap between the two halves of a balanced coil, using a large swinging coupling coil with almost as many turns as each half of the main coil. This was before anyone apparently had thought of the idea of a small, 3 or 4-turn swinging link to a separate secondary coil tuned to resonance with a capacitor. There was plenty of information on link coupling dating from the early 30s, but nothing on the swinging link. I can't imagine why it took so long (till the late 30s) before anyone thought of swinging or rotating the link to vary the coupling, since the electrical principle of link coupling was already long understood, and the vari-coupler had been in use in receivers since the 1920s.

I would guesstimate that a HDVL type coil would need more like 8 turns on the link to work with a simple series capacitor to ground to tune the link and achieve high enough Q to fully load the final to a 50-ohm load.  Remember, with a parallel tuned circuit, Q increases is the C is increased and L is reduced, but with series tuning, Q increases with increased L and less C. Series tuning works best when the resistive load is low (50-100Ω), but parallel tuning works best with a high-resistance load(≥ 1000Ω).

Another suggestion for the 160m coil and swinging link: connect a series inductance to the hot end of the link, directly ground the cold end of the link, and wire the variable capacitor to ground at the opposite end of the series inductance.  Connect the inner conductor of the coax to the junction between the series inductance and the capacitor. This forms a simple L-network to step up the low impedance (< 20Ω in my case) to 50Ω. 3 or 4 turns approximately the same diameter as the link would probably do, although I would start off with more than that and tap down till I found the proper value. A smaller diameter coil with more turns could be used give the same inductance. Several thousand pf of capacitance might be needed for resonance at 160m. A fixed mica with a variable cap of several hundred pf in parallel for fine tuning would be the best bet.  Use as little capacitance (with appropriate inductance) as you can and still find a resonant peak and fully load the transmitter with the link about 85% into the coil.

[W0BTU]

For my fixed-link HDA and HDVL coils in my balanced antenna tuner, there are the same number of turns in the fixed link for 80 through 10 meters. Only the series link capacitor has to be switched.

Photos at http://www.w0btu.com/files/antenna/balanced_antenna_tuner.
  The SPDT switch in the foreground selects the series variable capacitor, and the SPST in the background adds a mica padder in parallel with one of those variables.
  The coil on the left is for 20 through 10, the middle coil for 40, and the rightmost for 80. (EDIT: Or maybe a fourth coil, for 10 and 15, is in the tuner. I forget. I haven't used it since 2003.)

The SO-239 connector in the background is the input from the transmitter to the series link capacitor. Not shown is the ladder line which I alligator-clipped to the coil.

I've used this tuner at several QTHs, and the only thing I ever had to modify was the point where I clipped the ladder line to the coil. Maybe I just got lucky. :-)

 I would certainly think that one would need more turns for a 160 meter link.

[KC2ZFA]

There's a very nice article on link coupling to low- (series-tuned) and high- (parallel-tuned) impedance loads: QST, October 1959, pp. 29-31.

[k4kyv]

Pulled that one out.  Nice article. You rarely if ever see anything like that in QST any more.

If you have a lot of scrap coil stock and variable capacitors lying about, trial-and-error would probably be just as quick, but then check your result using the calculations per the article.