Choosing a PXI digitizer
Richard A. Quinnell, Contributing Editor -- Test & Measurement World, 9/1/2005
A rising application for PXI systems is the creation of virtual instruments, a collection of hardware elements that can be linked in software to perform the function of many different desktop instruments. A core element of such virtual instruments is the high-speed digitizer. Recently, the PXI Systems Alliance (PXISA) released a Web seminar that can help test engineers choose a digitizer for virtual instrumentation.
The seminar, "PXI Instrumentation and Replacement of Conventional Instruments," originally broadcast on June 28, 2005, and now archived on the PXISA Web site, has two parts. The first is a brief introduction to virtual instrumentation using PXI. The second provides a more in-depth discussion on choosing a digitizer; it is presented by Richard Soden, product manager at instrument maker Acqiris.
 |
| Fig. 1 Synthetic instruments combine a set of hardware blocks with software that allows the set to operate as a variety of bench instruments. Source: Synthetic Instrument Working Group. Courtesy of PXI Systems Alliance. |
The basic idea behind virtual instrumentation is simple. Most bench instruments have many hardware elements in common, such as displays, control panels, power supplies, and signal-processing stages. A virtual instrument would have one of each hardware element (Figure 1) and would route data through the hardware set under software control to duplicate the function of a bench instrument. Different combinations of hardware elements and data paths would allow the single hardware set to function as any of an entire suite of test instruments.Banner specs: Good start but not enough
A key hardware component for making measurements is a high-speed digitizer. Soden cautions test engineers to look beyond what he calls "banner specs" when choosing a digitizer to ensure that the component will perform as expected. Banner specs are performance specifications that serve as a quick guide to choosing a product, but if you limit your evaluation to only those specs, you will get a false sense of the product's capabilities.
Three banner specs that Soden identifies for high-speed digitizers are resolution, sampling rate, and bandwidth. These specifications are useful as a pre-filter when selecting a digitizer, and if a product's banner specs do not meet or exceed application requirements, then no further evaluation is needed. If the banner specs do meet requirements, though, you will need to perform further investigation before making a final choice.
The resolution banner spec, for instance, is an indication of a digitizer's accuracy. Soden notes, however, that the presence of noise and distortion in the conversion process can erode accuracy, so he suggests that you examine the effective number of bits (ENOB) that the digitizer provides. You can calculate the ENOB with the equation below (Ref. 1); the signal-to-noise-plus-distortion (SINAD) ratio is generally a function of frequency and signal strength presented as a graph:
 |
Another banner spec is the digitizer's sample rate, which is an indication of its timing accuracy. But knowing the sample rate alone is not enough. To fully understand a digitizer's timing accuracy, especially if it uses a built-in clock, you also must study its clock accuracy and clock jitter.
The clock accuracy, or frequency drift, will limit the accuracy of frequency measurements the digitizer can provide. Clock jitter will disperse the incoming signal's energy randomly throughout the spectrum, where it will be manifested as noise, reducing the ENOB.
Another important sampling parameter is the time to conversion. If a sample is assumed to be instantaneously captured, then there will be a phase error in the reconstructed waveform. By knowing the time to conversion, you can correct the phase information.
 |
| Fig. 2 While bandwidth is an important specification for digitizers, engineers should look beyond the simple 3-dB number to examine how the digitizer’s gain varies with frequency. Source: Acqiris Technology. Courtesy of PXI Systems Alliance. |
Bandwidth is the third banner spec that Soden notes is often misinterpreted. When the spec is provided as simply the 3-dB point, you might assume that the gain curve is flat and then falls off linearly. As shown in Figure 2, though, the gain curve of a digitizer may not be flat, but it can have the same 3-dB point, and hence the same bandwidth banner spec, as another digitizer with higher gain fidelity.Looking beyond banner specs
The key to making an accurate evaluation of a digitizer, therefore, is to look deeper than the banner specs. As a starting point, Soden recommends that you evaluate several secondary specifications. These secondary specs are direct measures of signal integrity and measurement fidelity, both of which will affect the accuracy of the digitizer's final output.
Along with ENOB, clock accuracy, and clock jitter, the secondary specifications include:
- Time-to-digital conversion (TDC).Knowing when the sampling trigger occurs relative to the sample clock helps reduce phase-measurement errors.
- Gain flatness. Examining the digitizer's Bode plot gives you a much better measurement of the digitizer's useful bandwidth than the bandwidth banner spec.
- Signal-to-noise ratio (SNR). To be useful, the SNR should always be quoted with the input frequency range for which it is valid, the digitizer's full-scale voltage range, and the input signal voltage.
- Spurious free dynamic range (SFDR). SFDR is a measure of the strength of the first spurious signal that the digitizer generates, relative to the input signal. As with SNR, this specification should include the input frequency and full-scale voltage ranges along with the input signal voltage.
- Total harmonic distortion (THD). THD tells you the relative strength of the signal and the first several harmonics that the digitizer generates. Unlike many analog systems, however, the higher order harmonics in digitizers may be stronger than lower order ones. Look to see how many harmonics are included in the calculation when making comparisons.
- Differential nonlinearity (DNL). The DNL indicates the input signal difference that results in the output moving between two adjacent bit codes.
- Integral nonlinearity (INL). INL is a measure of the difference between the input signal that produces an output code and the idealized interpretation of that code. Both INL and DNL represent deviations from the assumed ideal mapping between the digital output value and the input signal. Errors here can affect measurements of signal gain and offset. Matching the INL and DNL of digitizers is particularly important when using multiple digitizers with interleaved outputs to sample high-frequency signals.
A more in-depth discussion of these secondary specifications and of high-speed digitizers in general can be found at the PXISA Web site (www.pxisa.org). A recorded version of the Webcast is available to registered users (registration is free) along with Webcasts on other topics.
| Reference |
|
