Optimum length of exciter - HPA interconnect cable


It is often reported that the power available from an external PA is affected by the length of the interconnecting drive cable. Some even propose a 'magic' length that is optimal on all amateur bands, or that ferrite chokes on the outside of the coax fix the problem.

This article explores the issue in a practical experiment.


A certain installed TS2000 - AL811H combination is known to have a moderate HPA input VSWR on 7MHz.

A quality VSWR meter was inserted between the exciter and HPA, the HPA  tuned up for maximum power into a dummy load at 7.1MHz. The HPA input VSWR measured was 1.7.

Note that the cathode current in a Class AB/B grounded grid amplifier (such as the AL811H) has substantial harmonic content. This input network typically reduces the transmission of the to the exciter somewhat, but there may remain sufficient harmonic content in the interconnecting cable to give some apparently anomalous VSWR readings as VSWR meters typically depend on single frequency excitation.

Various length interconnect cables were tried (including the inline directional wattmeter) and exciter power out set to 50W. HPA output varied from 350W to 500W for the different cables tried.

Power output is observed to vary with length of the interconnecting cable.

VSWR measurement shows that the input impedance of the HPA is not close to 50+j0, so the interconnecting cable will transform the input impedance of the HPA to some value that is dependent on the length of the interconnecting cable, but in no case will the exciter see its preferred load of 50+j0.

A further measurement was made using an inline 75Ω VSWR meter which read VSWR=2.5. The intersection of the two VSWR circles on a Smith chart shows the solution points are very close to 30+j0.

Checking the original design

A model was constructed using the RF Power Amplifier Tube Performance Computer of the AL811-H HPA with 572B tubes. The model indicates that in class AB2, peak fundamental anode current is 1.08A occurs with 88V peak drive. From the characteristic curve, peak grid current is about 0.1A, so cathode current is 1.18Ap at 88Vp drive, yielding a cathode RF load impedance of 88/1.18Ω=75+j0Ω.

The AL811H circuit shows a low pass pi section filter for input impedance transformation. The network is some distance from the cathode, and is probably connected by a short length of 50Ω coax.

Fig 1:

A model of the circuit was constructed in RfSim99 to check the sanity of the circuit (Fig 1). The capacitors are from the user manual circuit, the length of 50Ω line is a guess at 150mm, and the inductor is tuned for best VSWR at about 7.1Mhz.

Fig 2: Modelled S11

fIG 2 shows S11 over the range 7.00MHz to 7.35MHz. Return Loss is -S11, so Return Loss varies from 25dB at 7MHz to 20dB at 7.35, with a minimum of 32dB at 7.12MHz. Return Loss of 32dB corresponds to VSWR=1.05, so the circuit appears sane and should be capable of adjustment to quite low VSWR.

Simulation indicates that to obtain an input Z of 30+j0, the value of L is a little too low (~1.0µH), so a small increase in L should solve the problem.


A quality VSWR meter was inserted between the exciter and AL811H, the amplifier tuned up for maximum power into a dummy load at 7.1MHz, and the input matching inductor adjusted for minimum VSWR. The VSWR achieveD was  less than 1.1.

Various length interconnect cables were tried (including the inline directional wattmeter) and exciter power out set to 50W. There was no noticeable difference in power out with the different interconnect cables.

There is nothing wrong with the 40m design, the AL811H just needed proper adjustment of its input circuit matching (done without removing covers).

This might look like matching issues are just a matter of alignment, however:

  • the design of the AL811H is lacking for bands where one input filter is shared for two bands, ie all bands above 40m. Whilst the specs are cleverly crafted to specify the maximum input VSWR as 1.3 at resonance, a single filter cannot be simultaneously resonant on two bands, so at least half the bands above 40m will have POOR input VSWR; and
  • the input L network on 160m exhibits VSWR=2.8 and appears to not be adjustable in having no core in the coil.

So, of the nine bands the amplifier covers, four have poor input VSWR that cannot be fixed of cannot be fixed without ruining performance on another band. Now there is a coincidence, it can be aligned for reasonable performance on 80, 40, 20, 15, and 10m... though the input VSWR on 15m and 10m is out of spec... basically a pre-WARC design with new panel labels!

One solution for an amplifier with such poor input VSWR behaviour as the AL811H is a small ATU between the exciter and PA.

It is doubtful that a magic length of coax will solve problems such as this.

Common mode chokes

If common mode chokes, or coiling up the interconnect cable make a difference to power output, that suggests that there is significant common mode current, and that the chassis of the exciter and PA are not at equal potential. This is an undesirable situation and efforts should be made to reduce ingress of common mode to the equipment room.

Installing a common mode choke between exciter and PA allows their chassis being at different RF potential, and defies good engineering practice which would normally indicate equipotential bonding of the two.

Collins 30S-1

Collins wrote the following advice in their documentation of the 30S-1 HPA, a grounded grid 4CX1000 amplifier.

Input Circuits

Pi-network broad-tuned circuits and the interconnecting r-f feed line match the 50-ohm input impedance to the cathode impedance, which is approximately 100 ohms. The 20.5-foot length of cable (furnished) is necessary between the 32S-1 (or KWM-2) driver and the 30S-1 input circuits. This is due to the necessity of having an even multiple of 180-degree phase shifts between driver plate and power amplifier grid. The cable length and the 30S-1 input circuits together accomplish this. An even multiple of 180-degree phase shifts is necessary because modulation components cause a change in the resistive PA cathode impedance which is translated to a shift in reactive impedance at the driver plate. The shift in reactive impedance, at the driver plate, results in phase modulation of the driver and increases the total over-all distortion of the system. A 2.5-foot additional length of cable is furnished to bring the total interconnecting cable length to 23.0 feet for use with the KWM-1 as a driver. Drive power required for maximum legal input on SSB is 80 watts PEP. 

To explore this issue, a model was created in RfSim99 of the total interconnect path from anode to cathode of a rational design.

Fig 3: Modelled interconnect path

Fig 3 shows the modelled total interconnect path from the anode of an exciter to the cathode of the HPA. The design is not the Collins, but a theoretical design for a grounded grid amplifier with an cathode driving impedance of 75Ω (eg the AL811H) and a 100W valve exciter running on a 1000V DC supply and requiring a resonant load impedance of 4kΩ, and using a low VSWR 50Ω interconnect. The interconnect path includes 6.3m (20.5') of RG58C/U as per Collins recommendation.

The network is tuned to present a resonant load of 4000+j0Ω to the exciter anode at 7.1MHz.

Fig 4: Modelled S12 (Gain and phase)

Fig 4 shows S12, the gain and phase of the modelled network. The phase shift from anode to cathode at 7.1MHz, is 32.8°, which is quite as expected, about -120° for each of the pi networks and about -85° in the 6m of coax, totalling -325° or 35°. Phase shift would be 212.8° from anode to grid, quite different (147°) to Collins  "necessity of having an even multiple of 180-degree phase shifts between driver plate and power amplifier grid".

It can also be seen from Fig 4 that the phase shift is very frequency dependent, so a slight shift in operating frequency requires a retune to maintain the phase relationship, but in the long run, a large component of total phase delay (due to the coax) is directly proportional to frequency.

Fig 5:

Fig 5 shows the phase shift with frequency assuming that the medium Q high ratio exciter pi network has 120° phase delay, and the low Q low ratio exciter pi network has 120° phase delay. It delivers the Collins target (even multiple of 360° anode to grid) at only one frequency, 26MHz.

Individual choices about the level of loading of the exciter, design choices of operating Q of the exciter output network and HPA input network all have influence on the phase shift from exciter anode to HPA grid.

In any event, the fixed length interconnect cable provides a phase shift proportional to frequency and cannot provide a frequency agile solution to Collin's stated "necessity".

The whole solution rationale is even less applicable to other exciter designs, eg broadband push pull solid state amplifiers with octave low pass filter sections.

None of this is to doubt that a Collins 30S-1 might work better with specified exciters using specified interconnect cable, but it seems more a work around for a design problem, their explanation doesn't seem rational nor to apply to typical designs in general.

Other cases

The problem described above can happen in a configuration at any frequency, but VHF and above amplifiers can be a particular problem when they lack end-user adjustable input matching. In cases where available exciter power is marginal for the task, mismatch may prevent achievement of rated output power, and an adjustable matching device between exciter and PA may be warranted. Depending on the frequency, this could be a conventional T match ATU (or use of a exciter internal ATU), a purpose specific L match, or at higher frequencies, a three stub tuner. Such a device will not eliminate dependence on cable length, but it will facilitate maximum drive power to the PA.


Output power level was dependent on interconnecting cable length when the input impedance of the AL811H did not match the transmission line well, so causing a length dependent load impedance to the exciter. The variation in output is probably attributable to the load characteristic of the exciter PA, and interaction with the power control and VSWR protection circuits in the exciter.

The observed variation is power with interconnect cable length was a symptom of a problem, incorrect adjustment of the AL811H input circuit.

The lack of variation in power with interconnect cable length with VSWR was very low suggests there are no other explanations for the initially observed dependence of output power on cable length.

There is no 'magic length' of 50Ω line that will transform the initial 30+j0Ω input impedance or any other impedance different to 50+j0Ω to the exciter's preferred load impedance of 50+j0Ω.

Installing a common mode choke between exciter and PA facilitates their chassis being at different RF potential, and defies good engineering practice which would normally indicate equipotential bonding of the two.

Where PA input impedance cannot be optimised, an impedance matching device can be inserted between exciter and PA to achieve maximum drive power, but the dependence on cable length will remain.

Notwithstanding the cleverly crafted specification of the AL811H, it does not meet spec, and the spec bears little relation to user needs.


RF Power Amplifier Tube Performance Computer


Version Date Description
1.01 14/02/2009 Initial.

V1.01 14/02/09 15:14:59 -0700 .

Use at your own risk, not warranted for any purpose. Do not depend on any results without independent verification.


© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.