A New Approach for the Environmental Compensation of OCXOs, TCXOs, and VCXOs
(For downloadable version, scroll to bottom of page.)
Esterline Research and Design’s newest patent-pending Multidimensional Segmented Array Compensation (M-SAC) technology is a platform for improving environmental performance of oscillators. M-SAC technology's method of environmental compensation can be integrated into discrete oscillators, custom ASICs, or system-level hardware. The technology advances the state of the art in oscillator compensation by providing the user with a substantially superior curve fitting tool. Unlike other contemporary fitting methods, the user can define a maximum error limit and the M-SAC methodology will find a solution that meets that criteria. This holds true for single-dimensional error, such as frequency versus temperature, or multi-dimensional error, such as trim effect (frequency vs. control voltage vs. temperature). With the M-SAC compensation tool, data can be fit down to the noise level.
ERD’s M-SAC intellectual property is ideal for mass-production environments. Due to the segmenting nature, data can be fit concurrently with the test run so that minimal number crunching is required at the conclusion of the run, maximizing throughput. Solution times are only a few seconds for single-dimensional data and under 30 for multidimensional data. If used with accompanying ERD test hardware, test and solution time is optimized by providing simultaneous data acquisition on all channels and incorporating proprietary means for reducing errors commonly encountered in oscillator temperature testing.
As shown in Figure 1, this method has enabled us to realize DOXCO performance from a Euro-sized OCXO.
Figure 1: Compensation of Euro-size OCXO
What is M-SAC?
Traditional oscillator compensation schemes utilize single-polynomial fits of the entire data set. This methodology has improved over the decades from thermistor-resistor networks to modern 5th-order polynomial ASICs. These methods are ultimately limited because neither the crystals nor the function generators in the ASIC produce
The M-SAC technology provides superior curve fitting by breaking the solution into segments that can be fit to a much better degree. The algorithm analyzes the maximum allowable residual error and slope tolerances input by the user and computes a solution comprised of segments that fulfill the user criteria. To accomplish this, the M-SAC algorithm evaluates a bank of possible functions and selects the optimal function for solution density.
Figures 2 through 4 below illustrate the advantage that the M-SAC technology will provide to your compensation schema and that the technology clearly provides a substantial improvement to the state-of-the-art technology in use across the industry.
Figure 2: Fit and residual error with a single 5th-order polynomial. Basically the industry's state of the art.
The first M-SAC fit for this unit was run with a user-defined error limit of ±25 parts per billion (ppb). Three segments were sufficient to fit the data within this error limit, as demonstrated in Figure 3.
Figure 3: M-SAC fit and residual error with three segments.
Table 1: Solution Array for Figure 3
Table 2: Solution Summary for Figure 3
The M-SAC fit in Figure 4 was run with a user-defined error limit of ±1 ppb. The M-SAC algorithm produced a 25-segment solution and fit the data to its noise level:
Figure 4: M-SAC fit and residual error with 25 segments.
Table 3: Solution Array for Figure 4
Table 4: Solution Summary for Figure 4
Figure 1 shows the frequency-versus-temperature performance of a COTS 2.0 mm x 1.6 mm TCXO and the fit achieved with a single 5th-order polynomial. The inherent performance of the TCXO is approximately 967 ppb peak-to-peak. This is essentially the current state-of-the-art performance offered by most contemporary TCXO ASICs. As Figure 1 illustrates, the residual peak-to-peak error from this fit is approximately 170 ppb. The error reduction is 5.7 to 1.
The M-SAC technology allows the user to define a level of acceptable error for the fit. With a user-defined error of 50 ppb peak-to-peak, M-SAC resolves a three-segment solution with a peak-to-peak residual error of 42 ppb. This fit and residual error is shown in Figure 3. With only 21 stored elements, the technology achieves an improvement of 23 to 1 over the inherent performance.
The power of M-SAC is that the data can be fit down to noise, provided the user is unfettered by data storage constraints. Figure 4 shows the same data set fit with a user-defined error of 2 ppb peak-to-peak. The solution is within the specified limit but uses 25 segments, requiring 132 stored elements. This results in a theoretical
improvement of 483 to 1.
M-SAC technology is the only oscillator-fitting schema that allows the user to define the residual error limits of the result. Advanced tools also allow the user to define slope limits, which gives the user more control and precision over the fit and resulting solution.
OCXOs are the stability “kings” of the oscillator world. With frequency-versus-temperature errors in the low-ppb arena, they provide superior, embeddable timing stability. And yet, even the resulting residuals from OCXOs have slope and shape, making them candidates for a secondary compensation platform. The M-SAC methodology is excellent for creating this secondary compensation, which can grant double-oven stability to a single oven unit across a wider temperature range.
Figure 5 depicts the frequency-versus-temperature performance of a Euro size (1.42” x 1.07”) OCXO with its inherent error as well as performance after the M-SAC compensation. The +/-0.27ppb of error over the temperature range of +85 to -55 C is a stability unachievable even with double-oven technology.
This is the best frequency-versus-temperature performance of an OCXO across this wide of a temperature range available by over an order of magnitude!
Figure 5 (For reference only, identical to Figure 1): Inherent versus M-SAC performance of a Euro OCXO
The stability and size of TCXOs have improved over the last 20 years, leading to their wide use in the industry. Standing out proves difficult in this crowded market because most manufacturers are competing with comparable products.
However, ERD’s M-SAC technology is poised to disrupt the TCXO market entirely. Utilizing M-SAC compensation on a TCXO, OCXO frequency-versus-temperature stability can be obtained with under 100 mW of power consumption. Figure 6 below shows the frequency-versus-temperature performance of a COTS 1.6 mm x 2.0 mm TCXO before and after M-SAC compensation. The TCXO with secondary M-SAC compensation provides exceptional frequency-versus-temperature performance, which can improve the residual error and slope by almost two orders of magnitude.
Single-oven OCXOs commonly have a significant warmup period before reaching stabilization and can often consume several Watts to maintain optimal performance over temperature. The M-SAC compensated TCXO operates at under 100mW and reaches stabilized OCXO performance in less than one second.
Figure 6: TCXO performance before and after M-SAC compensation.
Figure 7: Same data set, scaled to show compensated frequency-versus-temperature performance.
(Note: The inherent oscillator performance is almost entirely off the scale!)
Another product primed for market disruption by the M-SAC technology is the VCXO. These oscillators provide extremely wide pull, often in excess of +/-100ppm and are usually uncompensated for temperature. They have many applications and are often in phase-locked loops with OCXOs. Using the M-SAC’s multi-dimensional qualities, a wide-pull VCXO can be transformed into a wide-pull TCXO with negligible trim effect and excellent linearity.
Figures 8 and 9 below show a COTS 20MHz VCXO’s inherent linearity and the M-SAC compensated linearity respectively.
Figure 8: VCXO inherent linearity performance.
Figure 9: VCXO with compensated M-SAC linearity.
Trim Effect Compensation
Trim effect is a skewing of the frequency vs. temperature performance of an oscillator caused by pulling the frequency with the control voltage. This skew can commonly result in frequency-versus-temperature deviations that far exceed the specification (at control voltage extremes). This phenomenon is inherent in all oscillators with pull but rarely addressed.
Figure 10 depicts the trim effect of the VCXO once compensated for frequency versus temperature at center control voltage. Figure 11 shows the trim effect of the VCXO after M-SAC compensation.
Figure 10: Trim Effect Uncompensated
Figure 11: Trim Effect Compensated With M-SAC Technology
The M-SAC algorithm has turned a VCXO with +/- 15ppm temperature stability and +/- 12ppm of trim effect into a device that is +/-0.5 ppm across temperature and trim. The linearity was improved from 2.35% to 0.068%. M-SAC is an unrivaled method for trim effect compensation. It can reduce many systems that currently require phase locking a VCXO to an OCXO for stability to just the ultra-stable VCXO we have created.
Those with experience in the oscillator and RF industries are all-too-familiar with the problems encountered with crystal oscillator integration at the system level. The oscillator manufacturer’s production testing is rigorous; however, the conditions encountered in production testing often do not match the user’s environment. Many simple factors that seem harmless can have a large impact on oscillator performance, such as airflow, orientation, heat wicking to the PCB assembly, etc. When these problems occur, it can result in the customer and oscillator manufacturer spending time and resources in correlation testing and troubleshooting to isolate and identify the culprit and implement some mitigation strategy: usually some change to the production testing on the oscillator manufacturing side or costly redesigns of the product to reorient the oscillator in the package. These problems are a headache and money sink for manufacturer and customer alike. It can ultimately lead to loss of the client for the oscillator manufacturer if a mitigation schema cannot be deduced.
M-SAC provides a means for the user to integrate the oscillator in their system and then characterize it in the final environment. The algorithm can be used to then compensate the oscillator to the user’s specifications or down to the noise level of the data, whichever comes first. Even with no environmental issues from mounting or airflow, unforeseen aspects of the customer circuit that the oscillator manufacturer cannot know may come into play. For instance, the power supply for the oscillator's input voltage and control voltage will most likely have a temperature coefficient. This change in voltage over temperature will directly result in a change in frequency over temperature that can be detrimental. Figure 12 depicts a fictitious simulation of error caused a 50 ppm/°C temperature coefficient on the control voltage regulator.
Figure 12: Simulated post-compensation skew imparted by lumped system error.
By compensating using the M-SAC technology in the final environment, all lumped elements contributing to the degradation of the oscillator's frequency versus temperature response are virtually eliminated.
(Note: High-level systems may already contain the hardware necessary to implement the M-SAC technology. In such cases, OCXO performance can be obtained from the footprint of a TCXO.)
Esterline Research and Design’s newest patent pending-technology provides a great opportunity to oscillator manufacturer and user alike. For manufacturers, the ability to offer stabilities far superior to the current state of the art make M-SAC a wise choice for augmenting your product line for competition in the crowded market place. Your competition will be unable to touch your products in terms of sheer stability, performance versus power, and new classes of products to simplify user’s systems. For the user, the M-SAC technology allows you to compensate the oscillator in the final environment. This will allow for compensation of lumped errors that would normally affect performance. It also allows for the correction of problems that can be caused by orientation, airflow, and other deleterious effects.
This technology allows the user to define a desired outcome. No other product or method provides a means for a user-defined solution. The only factors limiting M-SAC compensation are space available for data storage and the noise level of the data.
Esterline Research and Design’s goal is to advance the state of the art in time base stability and environmental performance. We also license the technology to oscillator manufacturers and RF users. Our engineering team has over 65 years combined experience in crystal oscillator design and manufacturing. We have designed systems for the test and programming of the M-SAC methodology and hardware to facilitate integration into virtually any product or system. By utilizing our extensive experience, we can help licensees integrate the technology into their products and systems with minimal impact on time and cost.