THEORY of OPERATION for DANA'S MULTI-PHASE HF EXCITER INTRODUCTION: The HF system can be thought of as a 6-dipole phased array antenna system, with each dipole being driven by a separate transmitter. The correct drive phases for the antennas are generated at low power levels by the "multi-phase HF exciter", the very subject of this document. It was anticipated from the outset that the drive phases to all the individual transmitters will need to be customized (tweaked with respect to nominal values) in order to accomodate at least the following influences: > Length differences between all the transmission lines. > Phase mismatches between the different transmitters. > Residual resonant frequency mismatches between the different dipoles. > Cross coupling between the different dipoles in the array. > etc ... The essence of the scheme is to split the RF output of a single external RF signal generator six ways, passing each signal through its own variable phase shifter, with all to be set manually via a computer interface. Except where otherwise noted, the following discussion will be confined to a single phase shifter. Each phase shifter will permit setting both the phase and amplitude of its output, and must operate correctly over at least the frequency range of 5 to 8.5 MHz. BASIC PHASE SHIFTER THEORY: The heart of each phase shifter is a straightforward quadrature modulator, fed with I & Q DC levels obtained from a low-speed DAC on the same board. The DACs get their settings from a computer via a serial bus, and the phase and amplitude of the modulator's output can be varied anywhere over the IQ plane simply by setting the I & Q voltages as follows: I = ampl * cos(theta) Q = ampl * sin(theta) CIRCUIT DESCRIPTION (sans justifications & rationales; see later): Note: The phase shifters are configured one per ECB, so there are six ECBs in the chassis. What follows apples to each phase shifter. For the following description, refer to the schematic: 'hybrid_6-phase_synth_for_HF_VER_2.pdf' The quadrature modulator comprises mainly two mixers (M1 & M2), a quadrature L.O. splitter (QS1), I & Q output amplifiers (A1 & A2), a 7-pole lowpass filter (L1, L2, L3, C16, C22, C17, C18, C19, C20, C21, & C23), and a 0-degree combiner (IS1). In addition there is the usual assortment of fiddly little components (blocking capacitors, attenuators, etc). In addition to providing needed gain, the RF amplifiers (and the associated RF attenuators) help to provide accurate RF terminations for the mixers and for the input ports of the combiner (and thus for the lowpass filter). Note that the I & Q signals are fed in to the mixers' IF ports via fairly high-value resistors. I have learned over the years that in very low frequency modulating signal situations like this a current-source feed (or approximation thereof) leads to some useful improvement in linearity and reduction in carrier leakage. The DAC circuitry which produces the I & Q signals for the quadrature modulator is pretty conventional, straight out of the application notes from the manufacturer. IC1 is a convenient dual DAC in a single package. On the analog side it gets its 10-volt reference from IC4, a precision reference chip. IC2 (sections a & b) provide the necessary low-Z loads to the DAC sections and deliver voltage outputs, with the help of the feedback resistors that are part of the DAC. Sections c & d of IC2 are used to add in an additional DC offset from the reference so that the effective DAC outputs can swing both positive and negative. Thus a mid- range digital value will yield zero DC output regardless of the reference voltage, which is necessary because this design relies on turning the reference voltage on or off to enable or disable the RF output. Q1 & Q2, along with associated resistors serves to turn the DAC reference voltage on or off under digital control to switch the RF on or off. IC5 (sections a, b, c, & f) drive the digital control pins for the serial DAC. With the exception of the chip select signal, all the DAC signals and the reference on/off controls are driven in parallel for all the ECBs. But perhaps the filter networks ahead of the IC5 sections need some explanation: Consider, as an example, R72, R73, and C71 at the input of IC5_b. The passive parts form an RC lowpass filter with a cutoff of about 11 kHz. This overall system was designed to drive six 100kW transmitters, each of which is connected to a radiating element some 1100 feet away by 3-inch diameter Heliax transmission line. I was (and still am) concerned about the possibility of enough RF getting back into the transmitter building to disturb the operation of the DACs (or the controlling computer in a separate chassis) and make things go crazy. The 11-kHz lowpass filters here are to strongly attenuate RF before it gets to the inputs of IC5. Each IC5 section is a so-called "Schmitt Trigger", whose input hysteresis discourages flipping back and forth from noise or RF on its inputs, and which also serves to greatly sharpen up the now slow transitions coming out of the filters. ADDITIONAL COMMENTS ON THE DESIGN: > Rationale for selection of phase shifter methodology and the particular style: When the need for individually-adjustable phases was first realized, I began looking around a bit for a commercial product which could to the job. Muliple output synthesizers do exist, but careful investigation revealed that most could not generate synchronized outputs in the sense that when started the output signals would have a known or fixed phase relationahip. And indeed, the very few options we did find, by now rather late in the game, were still a bit unclear on this point and were very expensive for the frequency range covered. I also designed a DDS-based unit using a collection of AD9958 DDS chips, which feature two independent DDSs per package, including provisions for synchronizing the two in each package as well as for synchronizing across multiple packages. There ensued a lengthy period of working with this device type, communications back & forth with ADI, etc, but in the end we were unable to make this work due to difficulties in interfacing to and programming the AD9958s. By this time we realized that we were running out of time and something had to be done, pronto! At this point the idea of using quadrature modulators occurred to me. Present technology using cellular base station parts provides excellent accuracy at low cost- however, parts that will function as low as 5 MHz are rare. Linear Technology had one type, however, which suited nicely. I designed a test circuit and commenced laying out an ECB, but only then realized that the package was so tiny that IMHO we at the AO could not handle the task of mounting these things onto a board. So, I chose the build-it-from-a-kit approach, and designed a version which used discrete mixers, hybrids, etc, from Mini-Circuits Labs. While the phase accuracy would not be nearly as good as that obtainable from modern IC modulators, I felt it would be good enough for the job. In our situation we are relying on manual adjustments to walk the feed- point phases into the correct relationship in the face of unknown influences tending to disturb said relationships, so I felt it would be adequate as long as the quadrature modulators were well enough behaved that phase adjustment slopes were strictly monotonic and free of gross slope variations. The basic un-tweaked behavior of the proposed kit- form quadrature modulators easily meets this criterion. That is, if one were to command a particular channel's phase change, he could depend on the resultant actual change being in the expected direction and of about the expected amount, so that the process of homing in on the desired feedpoint current phase would not be unduly difficult. > Issues with the RF ON/OFF switching: As mentioned, this design and ECB layout were accomplished in quite a rush, and I overlooked the need to insist upon having complete system specifications before completing the design. In particular I was unaware of a desire for rapidly switching the RF on & off for some of the planned HF experiments. The present design relies on RF switching by turning the DAC reference supplies on & off for this purpose. In the static sense this approach works quite well; however the transition times are fairly slow (a few hundred usec) and the transitions are not clean. It now appears that this performance level may be inadequate for all experiments. Therefore I propose the following alternate scheme for the future: Leave the DAC reference supplies turned on continuously, and instead obtain RF switching by turning on (or off) the common RF input to all the boards. This can probably be managed very nicely with a FET RF switch such as the Mini-Circuits RSW-2-25PA and some driver circuitry. > Issues with RF level control upon system startup etc: As we worked with this multi-phase exciter system during the later stages of transmitter testing, it became clear that there were some significant problems. At heart, the problem was that any number of events could lead to the RF output level being set to maximum (or nearly so) against the user's will. This was in part a HW problem with the exciter's architecture, caused mainly by different parts of the digital interface being powered independently with no means of forcing them to be powered up in a safe sequence. It was also in part a SW problem, in that early versions of the control software were unaware of the problem. The fix as of this writing (10-28-2015) involves a partial SW re-write combined with bypassing the AC power switch on the exciter chassis so that it could not be turned off while the other stuff was still powered. It is our plan to move the last "stage" of the computer interface, an Arduino module, from a separate box into the same chassis on which the exciter is built.