Fundamentals of Rack Mount VXIbus Mainframes Abstract The following is a guideline for making cost-effective rack-mount mainframe selections considering mechanical features, power supply, cooling adequacy, system monitoring, and backplane requirements. Examples are given showing the usefulness of state of the art features. Mainframe features from several vendors are briefly compared. Introduction The heart of a VXIbus test system is its mainframe. Specific issues should be reviewed to decide the suitability of a particular mainframe to the needs of your instrument system. Some mainframe manufacturers do not specify their mainframes per the VXIbus specification which has requirements for comparison purposes. Therefore, the examples presented have tried to make fair evaluations with these VXI products. To indicate that a feature is not specified in the available marketing literature for comparison, an "Oh, no!" face appears in the table. Mechanical Features Experienced mainframe manufacturers know that room for routing system cables and mounting an interface panel or a receiver interface are key for a properly rack-mounted system. Servicing the power supply and fans without disassembling the whole mainframe is essential to lower MTTR and system down-time. System Cabling: A major concern is the space available for routing cables from VXIbus instruments to and from other equipment in the rack. Cables to instruments with signal connections in the front should be routed directly from the front of the VXIbus mainframe to the instrument. Likewise, cables to instruments with rear connections (or external instruments) should be routed horizontally underneath the mainframe through a cable tray. This requires the ability to route cables from module front panels either straight up or down or to the rear through cable trays. Wiring to an Interface Panel: Sometimes interface panels are used to connect VXIbus test instruments to the Unit Under Test (UUT) with specific connectors to speed test throughput. Such Application Specific Front Panels (ASFP's) require space between the mainframe and panel for cabling. Some vendors have optional ASFP kits. Mechanical hardware should be available from the mainframe vendor to allow a variable amount of recessing to accommodate specific applications. VXIplug&play Compatible Receiver Interface: The VXIplug&play Systems Alliance has developed standards to ensure simple receiver interfacing to mainframes from different vendors. The hole pattern in front of the mainframe must conform to this standard for VXIplug&play compliance with receivers from third party vendors. Both Virginia Panel and TTI Testron manufacture VXIplug&play interfaces to mainframes without restricting access to power switches, or other chassis functions. Mainframe Size: The rack depth of a mainframe should be small to conserve space within a rack. The unit should be lightweight enough to ease installation and serviceability. Different cable trays should be available for different applications. Table1: Comparison of Mainframe Mechanical Features ("Oh, no!" face indicates feature not listed in vendor's marketing literature) Power Supply One major advantage of VXIbus systems is the ability to have many instruments share one power supply, reducing size and enhancing system reliability. A mainframe must have enough capacity to prevent an over-current fault under worst-case loading. Data sheets for mainframes sometimes are unclear and unrealistic about the power capability of their products. Make sure that the vendor specifies the parameters required by the VXIbus consortium (e.g. dynamic current). Available Power vs. Usable Power: Mainframes are power-limited to a level below the sum of the peak current from each supply rail. Because of this, a mainframe's usable power rating is a better indicator of performance for a specific use than is available power. Available power is not a useful number on its own, yet some vendors emphasize it. Worse yet, they only specify available power since it is a larger number or because they do not know the usable power. All power supply specifications should be in effect from 0°C to 55°C or else ambient heating and self-heating of the power supply will derate peak power specifications. Specifying Dynamic Current: Vendors should specify mainframe dynamic current (IMD) which is the rated AC current of a particular voltage rail supplied by the mainframe over the frequencies from 20Hz to 1 GHz. Correctly specified VXIbus modules specify worst case module dynamic current (IDm). These currents must be summed for all modules and compared with the mainframe's IMD budget to verify system power integrity. Upgrading to Additional Power: Some digital and RF applications use a disproportionate amount of current on the +5V or +24V rails. In such cases adding an external, current-sharing power supply is a useful way to enhance power capability for these applications. Enhanced Reliability: Why is it important to have a powerful mainframe even if all of the power isn't required? The answer is reliability. If a mainframe's power supply is constantly running at full load, it will not last as long as one running at a moderate stress level (per MIL-HDBK-217E). It does not hurt to specify a mainframe with "too much" power (see Table 2). Table 2. Comparison of Mainframe Power Supply Features Cooling The VXIbus consortium has gone to great lengths to standardize the measure of cooling performance in a mainframe because of the big impact that excess heat can have on system reliability. To maximize the reliability of the components of a rack-mount system, adequate cooling must be available. To insure this, the mainframe must be designed to follow specific, industry standard air flow conventions. Cooling Operating Point: To compare mainframe cooling performance objectively and per the VXIbus specs, manufacturers must supply a cooling curve displaying the variation of air flow versus back pressure for the worst case slot. To compute the maximum power dissipation per slot, use the equation: AF = k * P / DT where AF = air flow in liters/second, k = 0.831, a constant, P = power dissipated in Watts, and DT = temperature rise in degrees C. With this information, finding the temperature rise of any (correctly specified) module in a mainframe is possible. For example, an attempt is made to cool a module rated for 70 Watts dissipation at 0.75mm back pressure. Air flow is read from two mainframe cooling curves (Figure 1). Curve 1's air flow is 5.5 liters per second, therefore: DT = k * P / AF = 0.831*80W / 5.5 l/s = 12.1 C. Curve 2's air flow is 7.5 liters per second, therefore: DT = k * P / AF = 0.831*80W / 7.5 l/s = 8.9 C. So, a better cooling curve means that VXIbus modules will run at lower temperature, enhancing system reliability (as evidenced by MIL-HDBK-217E). Table 3: Comparison of Mainframe Cooling Performance Positive Pressure: The efficient distribution of cooling air to modules within a mainframe requires the pushing of air through the module compartment (positive pressure) rather than pulling (negative pressure). This insures that properly filtered air is forced through the modules and not allowed to bypass occupied slots through empty ones. State-of-the-art mainframes are positively pressurized and have special air baffles in the chamber (plenum) below the modules directing air evenly across the module. Skirted, leak-proof air ducts are built into card guide systems to force more air flow through the module under high back pressure conditions (where most mainframes allow air to travel around the modules). Although not a VXIbus requirement, a reliable slot-blocking mechanism should be provided to reduce air flow in empty slots without cutting off airflow completely (slot covers may not be used during air flow analysis per the VXIbus specification). Slot covers are preferable to flimsy spring loaded covering systems which tend to break. Empty slot air flow blocking mechanisms should block no more than 80% of the air flow to keep from overstressing the fans. Proper Rack System Air Flow: Proper thermal management in a rack mount system requires attention to air circulation within the rack. In a typical test system, power supplies are at the bottom, VXIbus mainframes at table height, and any GPIB instruments are mounted above (see Figure 2). Rack-mount convention dictates taking air in on the sides and exhausting to the rear. This allows for a clear exit path from the rack for this hot exhaust air. The air flow schemes of many mainframe vendors are inconsistent with this convention, making it difficult to avoid the recirculation of hot air across VXIbus modules. The scheme is most frequently violated by taking air in from the rear instead of the sides. For comparison, air flow schemes are denoted as SR (side intake/rear exhaust), RS (rear intake/side exhaust), etc. (see table 3). Use of a "rear intake" mainframe in a rack-mount system will erode published cooling performance specifications. System Monitoring System monitoring is essential to verify proper system operation and to troubleshoot new VXIbus system applications. The integrity of VXIbus-based test and measurement hinges on the mainframe providing the services it is designed for. If the power distribution system or cooling system fails, test results may be rendered useless or modules may overheat and fail. Only the mainframe itself is in a position to monitor system parameters for the modules it hosts. Therefore, a user-friendly, message-based monitoring system is needed to verify test results. Versatile VXIbus systems require VXIbus system event monitoring to help in test program development and troubleshooting. With this, the mainframe becomes a true instrument within the test system. Power Supply Monitoring: To assure valid test results, it is necessary to monitor the current and voltage being supplied by each rail. Monitoring systems work by supplying monitored external analog signals, VXIbus alarms, and VXIbus IEEE-488.2 status register warnings. These can be combined to alert test personnel or the test program itself of power problems. For example, if a VXIbus rail current or voltage (in any combination) deviates from a user-specified tolerance, a status alarm may be sensed by the test program without any additional monitoring equipment. This information can be used to alert personnel to suspect test data. The most effective monitoring systems monitor both voltage and current. Undervoltage and overvoltage conditions are obvious problems. But if, for example, an open circuit condition exists (e.g. broken pin, blown picofuse), bus voltage may be measured by the monitoring system despite no current flowing to the module. Also, without current monitoring there is no way to protect a module that has failed and is drawing to much current. Having voltage and current monitoring has the additional benefit that total power may be derived (total power=SV*I for each rail) to detect an overpower condition. Therefore, it is crucial to monitor both voltage and current for all VXIbus power rails. Also, the integrity of a power monitoring system is maintained only when its power is derived from an independent power source that is not affected by malfunction of the main power supply (see Table 4 comparisons). Even if the main supply shuts down, the monitoring system should be able to send distress messages over an interface such as a RS-232 port. Finally, if the monitoring system fails or is removed, the mainframe should be able to continue to operate on its own without shutting down critical components such as the cooling fans. Cooling System Monitoring: It is necessary to measure air flow from each fan and the air temperature rise across each module directly to assure that no single VXIbus module overheats. These two measurements complement each other in assuring proper air flow. Cooling problems (e.g. bad fan, air blockage, components burning, etc.) within any single module cannot be detected unless each slot's temperature is monitored. Table 4: Comparison of System Monitoring Features VXIbus Event Monitoring: VXIbus event monitoring helps system integrators and service technicians to troubleshoot VXIbus related problems. If test failures are due to VXIbus problems, a time stamping feature can correlate these failures to invalid test data. Available mainframe technology allows for the monitoring, counting, and time stamping of events like VXIbus errors (BERR*), AC line failures (ACFAIL*), system failures (SYSFAIL*), and interrupt acknowledgments (IACK*). Useful features like these make sure that failure data is easy to track. Maintenance Event Monitoring: Serviceability is enhanced when a mainframe can keep track of its cumulative power-on time and notify the user when the air filter cleaning interval has been exceeded. This ability to track servicing not only aids in maintaining compliance with manufacturing standards such as ISO-9000, but also extends the life of the mainframe. Backplane Requirements At the heart of a VXIbus mainframe is the backplane. This part of the mainframe ties the entire system together by providing power and data interconnections for high performance instrumentation systems. Hidden features worth investigation include noise filtering, dynamic current, automatic configuration, backplane construction, and wireless mainframe design. Noise Filtering: Filter circuitry on a backplane is important in reducing switching power supply ripple and noise, high-frequency digital noise, and RF noise. Well-designed backplane networks employ tantalum capacitors for power supply noise reduction and ceramic capacitors for high-frequency digital and RF noise filtration. Dynamic Current: High-frequency signal sources (e.g. digital word generators, waveform generators, and signal generators) require stiff power rails that can source high-frequency current. In a conventional rack-and-stack (i.e. GPIB) instrument, this current is sourced directly from the electrolytic capacitors which form the output filter of the instrument's power supply. These capacitors have the low Equivalent Series Resistance (ESR) characteristic that decreases the power supply's output impedance to current sourced at high frequencies. As instrument formats migrate from rack-and-stack to VXIbus, current distribution is less localized and resistance and inductance are placed between the current source (mainframe power supply) and the load (VXIbus modules). The introduction of these parasitic effects makes it difficult for the mainframe to supply the required dynamic current. The VXIbus consortium addressed this issue by requiring mainframe manufacturers to specify how much dynamic current is available to each slot and specifying a measurement method. Backplane implementations must extend the power supply output filter by distributing low-ESR capacitors on the backplane near the connectors that supply this current to VXIbus modules. Table 5: Comparison of Backplane Features Automatic Configuration: Early backplane designs required mechanical switches or jumpers to daisy-chain the VXIbus Interrupt Acknowledge (IAK0-7*) and Bus Grant (BG*) signals across open VXIbus slots. Modern designs use active circuitry to make these connections automatically. Backplane Construction: The backplane printed circuit board must have low-impedance power distribution and proper impedance-matching of high frequency signals. The backplane must be mechanically stiff enough to handle many cycles of insertion and extraction without stressing signal traces. Wireless Mainframe Design: Wiring of mainframe fan, power supply, and monitoring devices should be reduced and replaced with a plug-in philosophy. This allows easy disassembly of the unit for repair and maintenance. In addition, less wiring reduces (copper) power dissipation and stray inductance, especially within the power supply wiring. As an example, current state-of-the-art mainframes are virtually wireless with a power supply module, fan module, and monitoring system that plug right into the unit. External Trigger Interfacing: GPIB instruments external to the VXIbus mainframe but located in the same rack are sometimes required to either trigger or be triggered by one or more VXIbus modules. One approach to dealing with this is to allow complete access to VXIbus triggers on the rear of the mainframe. VXIbus triggers can be sent and received and even delayed to improve trigger alignment. This results in significantly less cabling between the front of the VXIbus mainframe and other instruments. Summary Consider the ruggedness of a chassis: will it weather years of abuse or will it fall apart in a few years time? Consider the reliability of the components that make up the chassis: will they constantly run at stress levels approaching the operating limits or operate at moderate levels enhancing reliability? The mainframe is the centerpiece of a modern VXIbus test system. Attention to detail and long-term savings over short term cost will pay dividends when the correct mainframe for the job is selected. Remember that the mainframe is not just a box to hold your instruments. It is the heart of the test system and should be viewed as an "instrument" just like any other system component. TOC 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13