Consider this first example excerpt from a datasheet; a plot of the output impedance of a voltage reference with and without a capacitor. Identifying the manufacturer is not required as the issues that will be discussed are not manufacturer-specific problems. Similar issues can be found in data provided by most manufacturers.
While the graph of Fig. 1 seems straightforward, there are several issues that become apparent when one tries to make sense of what the data is saying. For example, the output impedance is a function of the reference output current, which is not specified. The dielectric of the 1µF capacitor is not specified, though the results would be significantly different for ceramic, tantalum and aluminum electrolytic capacitors. The voltage reference output current will also influence these parameters and, therefore, it should be included in the data and also considered in the circuit design. The most surprising issue, however, is that the impedance with the 1µF capacitor can actually be used to determine that the capacitor is not a 1µF capacitor at all, but a 3.3µF capacitor!
Since the bandwidth of the reference is approximately 4kHz, as seen by the peak in the output impedance, the capacitance can be determined from frequencies above this bandwidth. Arbitrarily selecting the point at 1 Ω and 50kHz results in:
There are several ramifications to this error. One is that with an actual capacitance of 1uF the bandwidth will be significantly higher, and so will the Q of the peak. This further means that the stability of the circuit will be degraded significantly more than shown. The PSRR and noise of the reference will be degraded significantly more with 1µF than with 3.3µF.
It should be noted that the output impedance measurement is very simple and can be quickly performed using low cost test equipment such as a vector network analyzer (VNA) with the appropriate signal injectors.1
In a second example (Fig. 2) also related to a voltage reference, the manufacturer makes a recommendation that unfortunately results in very poor performance of their voltage reference.
The addition of the recommended 0.1µF output capacitor results in very poor control loop stability even in the nominal case. The result of poor stability is poor PSRR, poor output impedance and increased output noise near 100 kHz, all of which degrade the precision of the reference. This is unfortunate, since 100 kHz is a common switching frequency for commercial power supplies. The stability and performance results will be further degraded by component tolerances.2
The first two examples both use voltage references, since they are very common and require very high performance in most applications, such as high speed DACs and ADCs. Similar issues also exist in linear regulators, which are topologically similar.3 Other common circuits include synchronous switching regulators and class D audio amplifiers.
A 2.4MHz switching regulator has ripple components beginning at the 2.4 MHz fundamental and including odd and even harmonics, depending on the operating duty cycle. Due to some modes of operation as seen in Fig. 5, some pulses are missed and the spectral content, as seen in Fig. 6, is far below the 2.4 MHz fundamental. In some operating modes, however, the spectrum content was measured to be as low as 22 kHz.
The corresponding spectral content seen in Fig. 6, indicates significant content at 600 kHz, only 6dBm below the fundamental 2.4MHz frequency.
In some operating modes, the spectral content is much lower in frequency, as seen in Fig. 7. The largest contribution to the spectral content is at 23.3 kHz, evident along with the higher frequency content seen in Fig. 6.
The output ripple may be much larger than expected and at much lower frequencies than expected as seen in Fig. 8. The low frequency content will also be evident in the input ripple current, affecting the converter or system EMI performance. At the input to the switching converter, the low frequency can also propagate to other parts of the system through the PSRR of other components.
Make Measurements Independently
In order to obtain the desired circuit performance it is essential that you first determine the critical parameters for your design. These parameters are often not included in the manufacturers data. Once you have determined the critical parameters you should make measurements independently from the measurements provided by the component manufacturer. It is preferable to make these measurements in-circuit using the actual circuit loading and operating modes. The failure to make these measurements can result in significant product issues, which can be very costly to correct, especially once the product is in production.
In the ideal case, it is best to make high fidelity measurements using the manufacturers’ Evaluation Module (EVM), in order to gain a general understanding of the performance of the part in different operating conditions and to verify the datasheet accuracy. Once a part is selected, it is important for the designer to test it again using the final circuit design, often in the breadboard or prototype state.
After the final design is constructed, it would be best for the designer to again measure performance to verify that the noise and printed circuit board characteristics, as well as the generally closer proximity of components on the circuit board, has not adversely impacted overall system performance.
- “New Technique for Non-Invasive Testing of Regulator Stability”, Steve Sandler, Charles Hymowitz, Power Electronics Technology, September 2011.
- “Are We Focused On The Wrong Reference Parameters”, Steve Sandler, Power Electronics Technology Magazine, February 2012.
- “Simple Method to Determine ESR Requirements for Stable Regulators”, Steve Sandler, Power Electronics Technology, August 2011.