GU-43B, GU43B, ГУ-43Б, Q-1P/41, Q-1P/42

1.5KW HF Amplifier

FRINEAR 1500

1.8–30 MHz, Input 50 ohm SWR ≤ 1.2

Click to Enlarge

COMMENTS:

While the circuit diagram provided the necessary basic information, care must be taken to use suitable rated components, adequate fancooling etc. Remember always that this is a high power amplifier with potentially lethal (HV) voltages.

This experimental design is capable of 1.5 kW output throughout the range 1.8–30 MHz, 5 W drive delivering 300–400 W. The input is "swamped" by a non-inductive 50 Ω control-grid resistor eliminating the need for tuned circuits to match the input circuit on the amateur bands. The simple screen supply incorporates protection circuitry comprised of a small 230 VAC/15 W domestic lampbulb as a constant current device. If the HV fails or the amplifier is not correctly tuned or is overloaded, the screen grid dissipation is limited to a safe value. If the pi-network output circuit is determined empirically for lowest screen current and maximum output, the screen supply is quite adequate and stable.

In practice the GU-43B and equivalent Polish Q-1P/41 and Q-1P/42 are very rugged; they withstood temporary overload during my tests: screen grid 300 mA, anode current 2 A and 100 W drive without 6 dB attenuator. The GU-43B is extremely stable in this design; no sign of oscillation even without in- and output loading.

Lowest screen-grid current occurs with anode voltage 3 kV≤ Ua ≤3.3 kV. With normal speech the current should flicker about zero mA (+/- 5 mA). With a carrier ("key down") the screen-grid current is about 20 mA positive.

Lowest screen-grid current occurs with anode voltage 3 kV≤ Ua ≤3.3 kV. With normal speech the current should flicker about zero mA (+/- 5 mA). With a carrier ("key down") the screen-grid current is about 20 mA positive

 

 

 

 

Q-1P/41 is a silver plated GU-43B; Q-1P/42 is a silver plated and ceramic GU-43B. Both types have a better amplification factor than the Russian GU-43B. They need 1.0 to 1.5 V more negative bias (e.g., extra 1–2 diodes in series) and deliver 100–200 W more output power.

 

 

Testing input circuit with 50 Ω dummy load.

Being a passive-grid amplifier, most of the input power is dissipated in a non inductive 50 Ω resistor. One might expect this arrangement to be frequency-independent; however, the capacitances of the control-grid, socket and associated wiring add up to about 100 pF which is only 55 Ohm (at 29 MHz) parallelling the 50 Ω resistor! This capacitance must be tuned out if what is adequate drive on 3.5 MHz is to produce full output on the higher-frequency bands. I do this with a dual-circuit.

Input circuit

Tubesocket with 50 Ω loaded broad-band input circuit; 3 Χ 15 nF (red) are removed for new components.

Eventual input circuits

Testing the design, left side: g2 supply

Screen grid supply

Action of the stabilizised screen grid supply, the lamp is used as constant current divice.

NOTE: Connect exclusively the (–Ug2) pole of the unstabilizised 470–580 V power supply to point "A".

Action of the QSK circuit

Hihg speed vacuum relay Siemens VR 311, 355 Ω/26.5 V

Control grid, relay and filament supply with single transformer

Pi filter output circuit

A toroid, self-shielding because of its low external field, facilitates a compact construction of an inductance. Before winding, several layers of Teflon plumbing tape must be applied to the core, to insulate it from the coil-windings. Another method of insulation is to cement two flat isolating washers (e.g. made from bare glasfiber board, figure 10b) on each side of the bare core. Apply a small quantity of super glue, possibly only a few drops, around the sides of the core. Work swiftly; the glue hardens quickly. The glue prevents the washers from moving out of alignment while the core is being preapred for winding. With a T200-2 core, the inner hole should be 28 mm diameter and the outer diameter 55 mm. With this last construction it might be even possible to use bare copper wire for the windings.

ANODE IMPEDANCE

Normally the anode- or output -impedance is known or can be calculated, or can be taken from a graph. However, hams are in the habit of (mis)using valves in other than normal ways, so available data don't apply. In the past I have built linears in a test circuit, following published schematics or design formulae and copying given values of components. After a while however, one will get an itch to experiment with this or that. Quite often in the following years, output was raised after tinkering with the output circuit. With simple means, the results were measured and tabulated, after which I tried to find a matching formula. I found my results to differ from the formulae and tables as found in handbooks of the years '60. In all probability, pure class B or C was then adhered to. Adjustment according to my "found" formula is a good directive for home-brewing linears that are meant for CW and SSB. The anode-impedance of an unknown amplifier's final stage with one or more valves, according to my findings is:

Ra = Va ÷ (1.87 × Ia), in which:

Ra = the (common) anode (or plate-)impedance, (in Ohms),

Ia = the (total) anode current at max. power, (in Amperes),

Va = the applied anode voltage, (in Volts).

 

One may dispute the value of the last decimal of 1.87, but please consider that it is the average result of many experiments. As it is, the formule worked very satisfactory for me.

 

CALCULATION OF PI FILTER

With my formula we find as a suitable anode load (Ra) for

Va = 3000 V/0.9 A:

Ra = 3000 ÷ (1.87 × 0.9) = 1783 Ω.

The anode-circuit must transfer energy and the circulating current amplification factor Q helps to suppress the generated higher harmonics. A loaded-circuit Q of 10–12 will meet most requirements regarding efficiency, suppression of harmonics and practical values of C and L. If we assume for the 80 m band

 

 

Q = 10,

then the loaded circuit-impedance becomes

Za = Ra ÷ Q = 1783 ÷ 10 = 178.3 Ω.

The tuning-C (=Ct) is mostly responsible for circuit-resonance, which occurs at the frequency for which

Zct = 178.3 Ω.

We recalculate to obtain pF's:

Ct = 106 ÷ 2πfZct, in resp. pF, MHz and Ω.

For 3.5 MHz this becomes:

Ct = 106 ÷ (2π × 3.5 × 178.3) = 255 pF.

This includes anode capacitance, wiring capacitance and stray capacitance! The circuit transforms the anode-impedance

 

1783 Ω down to 50 Ω,

giving an impedance ratio of

1783 ÷ 50 = 35.66

and a capacitance ratio of

√35.66 = 5.97.

The second C, normally called loading-C (=CL), has a value of

255 (pF) × 5.97 = 1522 pF.

Now we calculate the value of L for 3.5 MHz: Across the coil L is the series combination of

Ra + Rload = 1783 + 50 = 1833 Ω.

The loaded coil with a Q of 10 will, at resonance, see a resistance of:

Rs = 1833 ÷ 10 = 183.3 Ω.

Calculating for inductance gives:

L = Rs ÷ 2πf = 183.3 ÷ (2π × 3.5) = 8.34 µH.

It can be argued that the calculations with Ra = Va ÷ (1.87 × Ia) formula could be more exact. Just ask yourself what value you should assume for the added capacitance, caused by valve, wiring and stray capacitances...this uncertainty is much higher than the one caused by a somewhat simpler calculation.

Anode parasitic suppressor

NOTE: With some brands GU-43B the dimension from anode to base is about 5 mm too short for this

chimnee/socket, it is imposssible to sypply 12 V to the filament connector of the valve.

1st type socket

2nd type socket (see 12 × red 15 nF feed-through capacitors)

Click to Enlarge

For more information see www.nd2x.net/PA0FRI.html or the article in Dutch on this site.