Report: Summary of Response This report is a summary of the findings of the EVLA LO/IF PDR Review Panel and the responses by the Task Leader, based on a top level presentation of the design plans conducted on January 22 and 23 at Socorro and supplemented by a "delta" PDR on April 17. The purpose of the reviews was to answer 3 principal questions: 1. Are the top level performance requirements complete and adequate? 2. Have the correct design solutions been selected for study and development during the EVLA design phase: Are there important alternate solutions that are not being studied? 3. Has an adequate procurement plan been identified for the subsystem? Members of the Review Panel attending were the following: Dick Thompson, NRAO CDL Bill Brundage, NRAO ALMA Project Rob Long, NRAO ALMA Project Rick Perley, Project Scientist Jim Jackson, Hardware Systems Engineer Gareth Hunt, Software Systems Engineer Brent Carlson, Correlator Task Leader Terry Cotter, LO/IF Task Leader Steve Durand, Fiber Optics Task Leader Paul Lilie, Receivers/Feeds Task Leader Bill Sahr, Monitor and Control Task Leader A. System Requirements Specifications will be firmed up in several areas; such as, round trip phase, bandpath flatness (slope and ripple), polarization isolation, bandedges (3 dB), phase change (in ps and degrees), image suppression, goals for suppressing internal radiation, interface to front ends, and interfaces among preconverter and IF converter modules. Fractional dB ripple will be specified for wide bandwidths only. The flatness specification across the band of 3 dB may not reflect some degradation at bandedges. The flatness for a 2 MHz bandwidth will be 0.3 dB. The phase specification will be in picoseconds. Slope will be specified as maximum peak-to-peak variation per a given bandwidth. Frontend/IF interface *range* of total power and *spectral power* will be specified. System specifications have been allocated to subsystems. The 74 MHz specification for flatness and bandwidth will be resolved with the system specification. Suppression of images from frequency converters are not expected to be a problem because of high side injection of the LOs, but will be specified anyway. Thermal noise and large (10 J) signal inputs are not expected to impact stability as long as the system remains linear. "Data invalid" will be replaced by a flag included in the digital data. Uniformity of connectors is referred to systems engineering. PCAL has been removed. Fixed gain blocks are recommended for FE. Range of power levels for different observing bands needs to be better understood to minimize range of attenuation required in IF. Adding attenuation hurts S/N and headroom; the impact needs to be better understood and quantified. 30 dB step attenuator may be necessary for solar. VLBA observing will be provided for. MIB requirements are being developed in all areas. The number of distributed non-linear threshold detectors will be roughly equal to the number of band-reducing filters in the path. Schedule and manpower requirements are being defined more specifically. The number of ways to get to any given sky frequency will be defined. There is no plan to change correlator integration times to a multiple of 16.384 ms instead of 10 ms because 1) the sub-band transition-band aliasing attenuation will be good either way, 2) an advantage in short baselines where fringe rate is low would be offset by longer integration times which could cause imaging problems, and 3) hitting nulls in the fringe washing sinc function would require tuning to the natural fringe rate. Designs for Round Trip Phase are under development. SRAM memory will be replaced with memory that survives power outage. B. Preconverter modules The UX converter design is changed to provide for the block conversion scheme proposed by FE. The change dropped the number of switches required from three to two. Also with this change, the Ku-band converter would not just be a transition converter, but a permanent fixture. Three sets of power monitors are proposed: 1. At the input, one per polarization. Especially important where there is RFI. 2. Input of downconverter. 3. Head of sampler for AGC. Each monitor point will cost an estimated $1500 which drives interest in minimizing the number of points. Filters are to be located as early in the signal chain as possible. Two different specifications for peak-to-peak power are required, one for narrow band and one for wide band. The narrow band performance is primarily developed in the receivers, while contribution to the wide-band specs will be shared by the receivers, preconverters, down converters, and the interconnecting cabling. Spectral power specifications for preconverter modules at the lower frequencies need to take into account that the receiver band is narrower than the IF. The following are suggested: 4-Band ( 74MHz): -35dBm/2MHz P-Band (327MHz): -35dBm/50MHz* L-Band (1-2GHz): -46dBm/1GHz (compatible with dBm/GHz wideband convention) *However, all VLA documentation shows the P-band pre-mixer levels to be about -45 to -52dBm/50MHz IF. The MiniCircuits GALI amplifiers selected have a P(1dB) of about +18dBm. This yields a gain headroom of: For Pin = -35dBm, headroom = 53dB For Pin = -45dBm, headroom = 63dB Therefore, the discrepancy in P-Band receiver output power needs clarification in order to properly characterize the dynamic range and headroom in the 4/P-Band Converter. Input attenuation prior to the amplifiers can extend the headroom if the contribution to the system noise/SNR is acceptable. L-Band meets the 60dB headroom spec. Total power will be monitored in the L/S/C Band converter (which would also include the 4/P Band) and in the Ku-Band Converter. This would serve well for remote diagnostics in isolating problems occurring between the receivers and the Down Converters, and in checking for gain compression due to RFI. We will not be able to tell if a single diode goes bad from the monitor data. We will have to rely on data from experiments. One-third the system specification is reserved for the front end, one-third for the preconverter modules (T302,T303,T301), and one-third for the baseband converter (T304). The preconverter design includes a combination of low-pass and high-pass filters for an IF passband of 1090-1400 MHz, encompassing the RF spectrum from 66 to 376 MHz. Thus, 4- and P- Bands can be observed simultaneously, and their respective bandwidths could be expanded considerably. With the current 1024MHz LO scheme, the maximum bandwidth of P-band would be limited to 300 to 976 MHz, and image products could result. 1 V peak-to-peak (about 0dBm) input to the samplers is required for 4- and P- bands and will be provided. The resulting L-Band IF could also be expanded right up to 2GHz (from 1.4GHz), if needed, for a future receiver in the 400-970 MHz range, by simply changing the low-pass filters within the converters (about $150 each, or $600 per converter). Additionally, the 400-1000 MHz range receiver could be implemented via a separate, dedicated converter in the future. Components at L-band are relatively inexpensive. However, it would be problematic to lower the IF much below 1090MHz, corresponding to an RF frequency below 66 MHz, due to the 1024MHz LO. Solar attenuators for 4/P-Band exist on the VLA now (external to receivers) and will be included in the 4/P-band converter. The prime focus systems design is compatible with Phase II. The L-Band LO is 13GHz, which appears to present no threat to the L-Band RF or the resulting 11-12GHz IF. The 1024MHz LO is used to convert 4/P-Band to L-Band, and and could leak into the L-Band Converter via the IFs. We have provided for adding additional coaxial LO reject filters in the 4/P-Band Converter IF outputs, and/or adding 1024MHz LO reject filters in the L-Band Converter, if necessary. We may use a switch/relay in the 4/P-Band Converter to disconnect the 1024MHz LO from the mixers (terminate them into a resistive load) when this preconverter is not in use to further isolate leakage power. On the other hand, the IFs from the 4/P converter will be terminated by the L/S/C Band Converter selector switches, offering another level of isolation when the 4/P-Band Converter is not in use. Since the 1024MHz fixed LO is being generated independently from the 1st LO Synthesizer, fairly good isolation should exist between these two LO sources. New isolator specifications will determine if more isolators are required. The effects of frequency changes larger than 2:1 are being studied. The image rejection specification is currently -25 dBc but we intend to move it to -30 dBc. It is not known yet if harmonic images are a problem, but all images will be required to meet specification. 2-stage filtering will be required for a 2 MHz sub-band. Both stages will be FIR filters with (hopefully) ~1000 taps each. The data will be requantized to 4 bits after the first stage, before going into the second stage. FIRs are linear-phase filters with configurable bandpass characteristics. The design goal for polarization isolation is 80 dB; the minimum isolation permitted will be 70 dB. For LO leakage specs, 60dB LO isolation would be sufficient for the Correlator system. In summary, IF-IF Isolation: 70dB or better LO Isolation: 60dB or better The LO used in each converter passes through a power splitter, a high isolation amplifier, and another power splitter for driving the mixers. The isolation offered by these three major components is at least 20dB each, or about 60dB total. The RF-LO isolation would be about 80dB, due to the added mixer isolation. The input/output band-edge overlap for the C to X conversion will have 30 dB isolation goal, 20 dB minimum. Flatness will be specified per GHz. Specification of 1 dB per 200 MHz does not extrapolate well to 2 GHz. Band overlaps such as the one between input and output on C band converter will be addressed by the Front End Task Group. The fixed LO for 75 MHz should be changed for future wider band tuning requirements. C. Band switches VSWRs are reasonable considering the frequencies involved Isolators are being provided on the inputs and outputs of modules to reduce ripple. Isolators will be selected for stability with time, temperature, and gravity loading. Rack locations will be located to minimize lengths of high frequency signal runs. Attenuators may be required if Total Power (TP) specs cannot be achieved across the wide range of front ends. Given the limited number of COTS amplifiers available at the higher frequencies, it may be necessary to use attenuators to equalize the TP going in to the converters from the various front ends. For the band switches, the transfer switches, and the LO switches, switch updates should be adequate. Power levels to the various front ends or from the various front ends will not cause damage with these switches. There may, however, need to be fail safe switches within some of the modules to eliminate possibilities of damage. There are no problems with flatness of the transfer switches, but there is a problem with isolation and the spec will be changed. D. Baseband Converters During the "delta" PDR, calculated values for system-wide SNR and headroom were presented. In some cases, an SNR less than 20 dB was shown which was a concern because 20 dB SNR is thought to be the minimum requirement at the input to the sampler to avoid losing data. New calculations adding gain blocks at the input to T303 and T302 converters now show adequate SNR. As a result, the noise increase through the system is less than 1% for all bands. Insufficient headroom may still exist in the Q and Ka band receivers at the post amp and mixer which is a Front End issue. If so, it may be necessary to shuffle gain between these receivers and the T303 converter module. The headroom definition used is the delta between signal level and P(-1%), where P(-1%) = P(-1dB) - 12dB. Regarding the question of intermod levels relative to signal levels, a calculation shows that intermods are 104 dB below P(-1%) or 84 dB below carrier if signal kept 20 dB below P(-1%) level. The time constant of the power detectors will be investigated to determine if it is sufficient to blank data in the correlator. The availability of total power detection over the full 8 - 12 GHz bandwidth will assist techs in performing diagnostics independent of the correlator. Digital variable attenuators which are switched during observing are anticipated to introduce +/- 5 deg of phase error up to 20 dB of attenuation and +/- 10 deg of error up to 32 dB of attenuation; however, the attenuator will be used open loop in normal observing where no switching occurs and, therefore, no phase error will be introduced by attenuator switching. There may be special cases such as solar observing or observing with intensive RFI where closed loop operation is required, so we will seek attenuators with less switch-induced phase error. Monotonicity is guaranteed by the attenuator vendor. The baseline plan calls for use of the ALMA power detector, but when we require continuous ALC, we may not require as many bits. The granularity required for attenuation is currently unknown, but fractional dB is available. Assuming no change in input level, switching in 32 dB of attenuation in the downconverter produces a 15 - 20 dB degradation in SNR. For the case of significant attenuation being introduced in response to a significant increase in input level, no SNR degradation is seen. Reasonable software vs. analog feedback control modes for attenuators will be addressed in software. ALC to hold constant voltage and equalize among the 4 downconverters. Spacing questions, i.e. performance in the presence of sudden, strong RFI, will be addressed during testing. A question was raised, "is 1 GHz sampled bandwidth at 8 bits adequate for all observing or is 2 GHz sampled bandwidth at 3 bits necessary?" The response is that the project will stay with the baseline plan. There are no plans for multiple band observing except for 4 and P band where dichroics are unnecessary. No dichroics are included in the base line plan. E. Samplers Whether or not to use ALC of the signals at the sampler in the presence of interference is not clear. For the 8-bit sampler it may be best to keep the gain constant. It also depends on the headroom. It may be that the question effects only the control software, not the hardware. After actual instrument performance is measured, we can choose between analog ALC and software ALC/fixed gain. 100w dissipation is a worst possible case; the actual dissipation is probably closer to 50w. The module design distributes dissipation throughout the module and provides for adequate air flow for this amount of power. The baseline plan does not provide for PCAL. F. Transition Converter ALC is expected to respond adequately to concerns about RFI SSNR and headroom. The old front ends will be compatible with the new converters. As a worst case, power levels may be slightly higher and require attenuation in the path. G. DDS (Direct Digital Synthesizer) The preferred name, DDS, is used. Other names are the Fringe Generator and Fine Tuned Synthesizer. The design guarantees that frequency offsets are as stable as the hydrogen maser with respect to time. In response to concerns about inadequate tuning steps in the preliminary design, a 4096 synthesizer was proposed during the "delta" PDR. The scheme calls for 500 MHz steps in LO1 and 10 MHz steps in LO2. The 4096 synthesizer will reduce multiplication rates in the DDS. Cost will be $2500 for each 4096 MHz synthesizer; two per antenna. H. Central Reference Generator In response to questions about dump timing, requirements for 10.24 ms timing or timing over some other interval will be investigated. Currently, transition timing is dominated by the waveguide cycle, 52.083 ms (19.2 Hz). The design calls for fringe rates and phase rotation that exceed the requirements. Integral reference generator cycles would also require an integral number of fringe cycles. Since the fringe cycles are a function of baseline and time, integral reference generator cycles are not considered practical. Hard specifications for peak-to-peak phase variations are in the Science chapter of the Project Book and need to be translated into engineering requirements. Time-of-day will be provided at the antenna. Currently, the IAT clock synchronizes the online system to "time-of-day;" but the LO/IF design will provide any timing signals required in the EVLA design, to include 128 MHz and 1 PPS for the correlator, and 1 pulse every 10 seconds for VLBI formatter. Adequate spare frequency ports will be available for the NM Array. I. Antenna Reference System Frequencies will not be divided at the antenna because the continuous re-setting of the dividers would cause blanking of references through the duration of the re-set pulse. As a result of this change, a clean-up loop will be required for the 128 MHz at the antenna. If we can keep the antenna timing signals to a minimum, we will consider using the ALMA design (although we are still studying it). The ALMA design, as we see it, will require a PLL for each timing signal sent to the antenna. The need for additional clean up loops to reduce degradation of phase noise in fiber is being considered. J. First and second LO System Proposals to provide a separate LO for each antenna or to generate separate frequencies at the CB are ruled out because of cost. The LO synthesizer design was changed during the "delta" PDR to permit 1 or 5 MHz steps. The additional resolution of the steps over the initial design presented in February will be needed during the transition. The new design is much the same as the one at Astron. The main difference is in the use of the DDS tuneable frequency synthesizers. The phase performance of the DDS relates to the "control loop" oscillator as it is called in the Astron design. With the 4096 synthesizer, high multiplication of DDS output rates does not appear to be necessary which in turn reduces concern about the phase performance. Testing is planned, never-the-less. The temperature sensitivity of the sampling mixer (phase v. temperature) is thought to be eliminated by biasing the diodes, but again, testing is planned. The principle characteristics of the synthesizer are short term phase stability (phase noise), long term phase stability and phase coherence between frequencies; the requirements are understood. In response to concerns about protecting the synthesizer when the module is swapped, it will be designed to power up at decreased output power. K. RFI issues. Shielding of the B rack, [and for that matter, the entire vertex room] will be studied. Careful attention to connector leakage, such as torque wrenches and conductive paint, is planned. Requirements for RFI suppression are now specified in the RFI Plan in the Project Book. The requirements were prepared by Dick Thompson and Bill Brundage based on IEEE Trans. Ant. Prop., AP-30, 450-456, 1982. The values are a suitably small fraction of the total power (power flux density) seen by the quantizers for 1 GHz BW/8-bit and 2 GHz BW/3-bit. Harmonic mixers are favored over comb generators because of stability and RFI. Filters are planned for bias lines. RFI testing for the French sampler prototype will be scheduled as parts of the design become available. The distribution of peaks in the DME (1 - 1.2 GHz) will be studied once calibration is available on the Environmental Monitoring System. Power monitors in the IF chain will check for linearity. A port may be added to the existing "W8" monitor to measure RFI to 12 GHz from the 0 dB sidelobe. The impact on RFI mitigation of using an integral number of fringe cycles in the correlator is a software issue. IV. Conclusions: The top level performance requirements for the EVLA LO/IF design are complete and adequate with the following important exceptions: Tuning range of LO synthesizer (addressed in delta PDR) Need plan for round trip phase. Need tighter specifications in several areas; details are listed. Need to define MIB (M&C) requirements. Need to define self-RFI suppression requirements. Maintainability testing needs more attention. Correct design solutions have been selected except for the following: The LO synthesizer tuning ranges which was corrected in delta PDR. The procurement plan calls for ordering long delivery items as soon as possible. Obsolescence was not addressed. For the panel Clint Janes