Alpha Monthly News Letter

The Alpha 77 and Alpha 87a are probably the two Alpha amplifiers that most changed the world of HF linears. The 77, with it's big single (or in a some caes double) 8877 tube is still loved by owners, and if one comes on the used market, it is snapped up quickly. The 87a, with its auto-tune features and microprocessor control was revolutionary. Around 1,500 of the 87a amplifiers were produced before the decision was made to replace it with a newer product. This decision to halt 87a production was driven by the realization that the technology in the control circuitry was getting seriously behind the state of the art. The hand-assembled through-hole technology resulted in a cost structure that was getting out of hand along with some parts becoming hard to find. Finally, the reliability of that many parts was going to be less than a more highly integrated design that would be possible with current technology. So about 15 years after the design for the 87a was started, we started in on a new design. One final thing that made the change seem like a good idea was that the price of the 3CX800A tubes had skyrocketed. A single 8877, on the other hand, could be had for about two-thirds the price of a pair of 3CX800As. In addition, imported versions of the 8877 were just becoming available, and promised an even lower cost of tubing an amp if they would prove to meet our requirements.

In starting the new design, we decided to keep what was still working well, and use that as a starting point. So we did not start with the classic blank sheet of paper. Once the decision to use a single 8877 was made, we looked at the 87a automatic-tune tank circuit. This works very well- we had switched to a new variable capacitor supplier, and the stepper-motors that drove the capacitors and the bandswitch were fine. So the tank circuit of the new amplifier, soon to become known as the 9500, was lifted more-or-less complete from the 87a. The biggest change was that we had to increase the voltage rating on the "tune" capacitor, since the 8877 likes to work at around 3.5kV or so, which was more than the 3kV or less on the 87a. The power transformer in the 87a was a unique design. As just mentioned, we needed more voltage for the 8877, and this transformer did not look like a good choice. We realized that with a few minor tweaks, the transformer from our tetrode line of amplifiers could be made to work in the new amplifier, by placing the screen supply (500V) in series with the plate supply (3kV), to get the desired 3.5kV. By adding a tap on the heater supply, and checking the engineering with our transformer manufacturer, we had something that was completely satisfactory for the 9500 and the tetrode amplifiers.

With these top-level choices made, the rest of the design was tackled. A major early objective was to radically simplify the wiring harness as compared to the 87a. After considering some alternatives, the decision was made to use the "Inter-Integrated-Cirucit" bus, commonly kown as IIC or I2C. This bus was developed by Philips in Europe for the same reason- to simplify the wiring of major domestic electronics appliances such as televisions. It is a nice solution- it uses two "wired OR" lines along with ground, and had transmission speeds that would be fast enough for our needs. It was not clear, however, if this could be made to work in the high-RF environment of a compact linear amplifier like the 9500. But a few "kinks" on the basic way I2C was intended to be used solved the problems. First, we use an auxiliary driver IC made by Philips, that allows much greater current drive than the standard I2C parts. This IC was originally intended to increase the length of line connecting bus components, but we realized it could also be used to allow more filtering on the lines to kill RF without overly distorting the I2C bus signals. Bus errors are still encountered, but these are dealt with by adding a so-called horizontal parity check to each block of data that is transferred. Blocks with errors are simply discarded. Since the amplifier "heartbeat" is 100 times per second, there is enormous redundancy in the data going over the bus every 10msec. Losing an occasional block results in inconsequential effects on performance, since the data is repeated 10ms later. The final piece of the puzzle was the right choice of ribbon cable to connect the bus to all the parts of the amplifier where it was needed. We chose a shielded form of ribbon cable and the shield is grounded wherever a connector is needed.

With a reliable way for parts of the amplifier to talk to each other, the major blocks were defined, and each one was designed around its' own microprocessor. There are five of these- the master controller board, and then four slaves: the power supply stack, the user interface, the stepper motor driver and a sound generator. Each one of these slaves can take care of it's own set of well-defined, logically-grouped tasks, and every 10ms exchange important data with the master controller. This architecture has numerous advantages. First, each part of the amplifier can be tested without the need to even have the I2C bus running. Also, the slaves can take action on their own, if this is appropriate. This can help the reliability and safety of the amplifier. For instance, the power supply slave can cut the power off for the entire amplifier if it fails to get I2C communications from the master for an extended period of time. Similarly, the stepper motor slave takes care of locating "zero" for the capacitors, reading the position of the bandswitch etc. It only needs to receive a pair of numbers from the master indicating the position of the tune and load capacitors, and it moves them to the new setting, without the need for further action from the master.

The 9500 was designed with fully remote operation in mind, and for this (and other) reasons, four output connectors were provided. This allows simplification in wiring of a remote station. Also, both USB and RS232 interfaces were provided to allow flexibility in connecting to a control computer. All front panel functions can be executed from a control computer.

The design has proven a success, combining the best features of the 77 and the 87a with a modern bus architecture and a distributed multi-tasking software system. This is a lot like a modern car, which has a gasoline or diesel engine with roots going back almost a century, but also has a dozen or more microprocessors communicating over a bus. Both are representative of the state of the art in their respective fields, driven by considerations of cost, quality, performance, safety and reliability.

Next month we'll talk about Mains supply.

Gordon Hardman, W0RUN