Two 4CX1000A’s in grounded-screen push-pull – an amplifier that did not work.

Describing this amplifier may seem a really silly thing to do, as it never worked well. However, someone else might learn from our mistakes, or even perhaps someone can tell us why it did not work.

Background.
A friend Paul G8WYI and I were are a radio junk sale when we were offered a pair  what appeared to be new  Eimac 4CX1000A tubes. Paul bought them, and I set about designing a 2 m  amplifier to use them. The first thing that became apparent was that the correct Eimac SK-800 series bases, with the screen decoupling capacitor, were very expensive. However, a design published in the 1991 ARRL handbook, A Legal-Limit 2-Metre Tetrode Amplifier, pages 31-57 to 31-72,  using a single 4CX1000A had avoided the need for such a base. The authors, K1JX and W1VD, had used a SK-800 series base with a damaged screen decoupling capacitor and grounded the screen for both RF and DC. They claimed that this avoided the need for the decoupling capacitor in the base and instead meant a cathode decoupling capacitor was needed. However, they argued that any reactance in this cathode decoupling capacitor would not make the amplifier unstable, but would act as negative feedback, and increase drive power.

We did not have any base, but I decided making one was not too difficult, having access to a good mechanical workshop at university.  A picture of the SK-800 base appeared in an Eimac publication, so making one seemed easy. In fact, making a base was a lot of work, and with hindsight is not something I would try again, but I did make a pair of bases that grounded the screens.

The data sheet for the 4CX1000A gives the following as typical operating conditions.

Anode voltage             3000 V
Screen voltage             325 V
Grid voltage               -60 V
Cathode voltage              0 V (cathode grounded for DC+RF)
Output power              1630 W
Zero signal anode current  250 mA

We proposed to run like this

Anode voltage              2675 V
Screen  voltage               0 V (screen grounded for DC+RF)
Cathode voltage            -325 V
Grid voltage               -385 V

Note that voltage differences between our proposed design and the recommended ones are the same. We have 385-325=60 V between grid and cathode, 2675+325 = 3000 between anode and cathode etc. We aimed to return the negative of the HT supply to the cathode and not the ground. Hence a 3000 V HT supply was needed, but the +V would be at +2675 V wrt ground and the -V at -325 V wrt ground.

With the DC sorted out, the next step was to design the RF deck. Having had a lot of success with a  well known twin 4X250B amplifier , I decided to base the twin 4CX1000A on this. The grids had two lengths of 6.3 mm (1/4″) pipe, inductively coupled to the input with a loop and a series capacitor. The anode line was in the shape of the letter U. with the tubes at each end of the U and the DC fed into the middle via an RF choke.

Switch on.

The amplifier was switched on and the DC operating conditions seemed okay. The zero signal anode current was close to the 500 mA expected (2 x 250 mA). It could be adjusted to be exactly 250 mA each tube, by taking the grid voltage very negative on one tube at a time to cut it off. It was clear the DC conditions were fine. It was also clear that the amplifier was stable no matter how we neutralised it – something necessary to get 100% right on the W1SL design, otherwise it took off and took the HV fuses with it. Our twin 4CX1000A amplifier was remarkably stable. So far so good !

Application of RF.
When RF was applied, the amplifier did a job of amplifying. In fact, with less than a Watt of drive, several hundred watts came out. However, it was soon apparent was that the amplifier started to draw grid current very early – by a couple of hundred Watts out, the amplifier was drawing grid current. By the time the output was up to 700 W, our grid current meters were at full scale ( 1 mA). I think we eventually ran up to about 800 W out, when the grids were taking a couple of mA each. Clearly the amplifier was non-linear and we were likely to destroy the grids of the tubes if we pushed it any harder.

The grid dissipation rating of the 4CX1000A is 0 W !!!

Despite a lot of effort, a lot of asking around, the amplifier still amplifies, but has never been used in a QSO, as we know it would be non-linear. Why it draws grid current so early is not known to us. We have thought the problem was one of incorrect loading, but have tried various methods of coupling the RF out from the amplifier with no success. No matter what we try, the positive grid current problem will not go away. If you have any comments on this, we would like to hear from you. by email  at drkirkby@ntlworld.com G8WRB is now making a conventional cathode driven YC156/3CX5000A7 amplifier and hopes he has less problems than this design that was unusual to say the least.