Calibrating a National Instruments 1 Digitizer System for use with M1 Oscilloscope Tools ASA Application Note 11-02 Introduction In ASA s experience of providing value-added functionality/software to oscilloscopes/digitizers over the past twenty years, we have observed that most hardware has certain quirks that can become an obstacle to making very high-grade measurements or to its usability. Historically, ASA has made the effort to discover these quirks (in most cases, we were the first to report them back to the manufacturer) and provide a reasonable mitigation within M1. In many cases, these mitigations were never even publicly announced, but they no doubt contribute to M1 s reputation as "the gold standard" vs the stock software you can buy from the manufacturer. Some of these anomalies have included: Understanding, detecting and managing changes in the calibration state of oscilloscopes due to room temperature variation Understanding and calibrating out brand/model/setting-specific platform noise from RjDj measurements Channel to channel dynamic drift and tools for managing it Accommodating "undocumented" behaviors in the software state of an instrument etc. Disclaimer We want to emphasize that there is nothing unusual about the existence of these "imperfect" behaviors, and what would actually be a surprise would be to encounter perfect hardware. We've never seen perfection in doing this for twenty years, and ASA just wants to provide our customers with tools that provide high quality solutions to these issues when they're present and which ultimately make the platform more useful to the customer. Further, the observations that led to this document are based upon direct hands-on experience with a small number of digitizers provided by either the manufacturer, our channel partners or our customers. We are not representing that our observations are either indicative of behaviors that will occur in every single platform of a certain type, or are complete across all copies of the platform across their service lives. The sole objectives are to note that ASA has observed something and to provide some tools and methods to reduce or eliminate the issues as a service to our customers. 1 National Instruments and NI are trademarks of National Instruments Corporation, Austin, TX USA Restricted Distribution pg 1 of 6 Edit date: 8/26/2011 11:52 AM
NI Multiple Digitizer Systems ASA has discovered a behavior in NI multiple-digitizer systems that could negatively affect channel to channel timing measurements when the channels are on different digitizers. ASA has also created a calibration tool and procedure whereby this behavior can be substantially mitigated. In order to realize the benefit of these mitigations, the user will need to perform the calibration procedure described below to maximize repeatability. It is recommended that this calibration be repeated every time the digitizer system is powered up, after the system has been allowed to warm up for at least 15 minutes. As is our practice when discovering such quirks, ASA has notified the manufacturer of the existence of this behavior. As of this writing (late Aug 2011), NI has confirmed the existence of the behavior and is working on finding a more permanent solution that will eliminate the need for this calibration procedure. Description of the Behavior Summary: When making a channel to channel timing measurement with channels on different modules, the inherent skew between the modules can sometimes change suddenly by one sample interval. The impact of this behavior can be seen in the graph below: Restricted Distribution pg 2 of 6 Edit date: 8/26/2011 11:52 AM
In the example above, we are sampling at 100 MSa/s (10 ns sample interval) and measuring a skew that is nominally -1.5 ns. We take several dozen acquisitions, each showing the expected Mean skew of -1.5 ns. But the noted behavior then manifests itself on a single acquisition, and we now see a measured Mean Setup time for that acquisition of +8.5 ns (-1.5 ns nominal plus one sample interval of 10 ns). The histogram contains the accumulated events from all acquisitions taken; notice the small outlier population circled at the bottom right which contains the events from the single acquisition that was affected by the behavior. Note that the outlier population could have shown up at -11.5 ns (-1.5 ns minus one sample interval) instead, depending on exactly how the behavior manifested itself. Note that the distribution and repeatability of the results from that acquisition are unaffected by the manifestation of the behavior. One of the channels is simply providing data to M1 that is offset by one sample interval from what was expected, causing the edge timing on that channel to be off by one sample interval, which causes every skew event in that acquisition to become one sample interval larger or smaller. In ASA s experimentation and testing, the behavior sometimes would not appear for several hundred consecutive acquisitions, but at other times could affect 15-20% of a similar number of acquisitions. For the majority of the time, the frequency of occurrence was in the 1-2% range. NI's recommended Best Practice for dealing with this behavior is to adjust a parameter that modifies the phase relationship between the synchronous cross-module trigger signal and the sample clock of the digitizer. By sweeping this parameter across its range and taking a sequence of acquisitions at each value, a value where the behavior either completely disappears or at least is greatly reduced in frequency can be found. This is a somewhat tedious procedure, and requires direct programmatic access to the digitizers to accomplish. Calibration Overview As the M1 digitizer user does not have direct programmatic access to the digitizers, ASA has created an App to automate this calibration. The NI Calibration App is available for download from the General Apps section of the M1 App Store at www.m1ot.com/m1apps. With default settings, the calibration sequence will take 15-20 min, during which you do not need to be near the system. The user does has control over two factors which can significantly influence the time required for the sequence to complete. The first factor is the number of acquisitions that are taken at each value of the phase relationship parameter being swept. The default value is 100; you can speed up the calibration by lowering this number, at the cost of a less precise understanding of how often the behavior manifests at each tested value and therefore a less precise final calibration value. You can also increase the precision (and the time required) by increasing the default value. Restricted Distribution pg 3 of 6 Edit date: 8/26/2011 11:52 AM
The second factor is whether M1 saves all of the calibration acquisitions it is taking. The default case mentioned above assumes that M1 does not save this data. Turning this option on will increase the calibration time to 30-35 min and require ~125MB of disk space. For the calibration input signal to the digitizers, it is suggested to use a pulse generator to provide a simple and clean clock signal. You should select a speed and voltage range similar to that of the signals that you eventually intend to measure for best results. For ease of setup, it is also suggested to use cables to transmit the signal from the generator directly to the digitizers, even if your real measurements will be made using probes. You will require a signal splitter and two cables of approximately the same length in order to split the signal into two modules. Calibration Procedure A single calibration sequence will calibrate the primary module against one other module. If you have more than two modules, you will need to perform the calibration sequence N-1 times the primary module with each of the non-primary modules. Only the first input channel (Ch 0) on each digitizer will be used as the behavior has no effect on timing between the channels of a single digitizer. Before starting M1, connect a single output of the pulse generator to Ch 0 of the primary module and Ch 0 of the module to calibrate with using the splitter as shown to the right. Select a speed and voltage range on the generator similar to that of the signals that you eventually intend to measure and enable the output if required. Now you should start M1. When the Digitizer Setup dialog comes up, select appropriate values for Triggering and Channel settings. If possible, use the values that you expect to use when measuring your intended signals. It is recommended to select an acquisition window length that provides a shallow memory depth (10-20kpts) for most efficient processing. For detailed information on the use of the Digitizer Setup dialog, see the M1 Help file. You should also determine the M1 channel number of Ch 0 on the module to be calibrated as you will need this information later. You can determine this from the Digitizer Setup dialog by hovering over the desired channel s icon; near the top of the pane on the left you will find the M1 channel number for that digitizer channel listed. This information has been circled in green in the image below. Restricted Distribution pg 4 of 6 Edit date: 8/26/2011 11:52 AM
Once your settings are in place, click on the NI Calibration button, which is circled in green above. The Digitizer Setup dialog will close, and M1 will shift into TestScript mode and run the NI Calibration App After a pair of description and warning dialogs, you will first be asked to enter the M1 channel number for the signal on the second digitizer. Enter the number that you determined in the previous step and click OK. Next you will be asked how many acquisitions to take at each test value of the phase relationship parameter. Change the value if you wish (and understand the implications based on the discussion in the previous section) and click OK. Finally, you will be asked if you wish to save all acquired data during the calibration process. Select your answer based on the discussion in the previous section. That s all! Now you just need to wait for the process to complete. M1 will place periodic messages in the status bar to notify you how far into the process it is. Once the process is complete, M1 will set Restricted Distribution pg 5 of 6 Edit date: 8/26/2011 11:52 AM
the optimal value of the parameter and provide a dialog noting what that value was. You can now Exit the TestScript Console and continue working in M1. Conclusion If you have a multi-module NI digitizer system and intend to make channel to channel timing measurements between channels on different modules, performing the calibration described in this document is highly recommended. Note that you may need to repeat the calibration each time the digitizer system is powered up as the ideal value of the phase relationship parameter can change. Restricted Distribution pg 6 of 6 Edit date: 8/26/2011 11:52 AM